Figure 1 illustrates the ESS concept. ESS is effective only for a product with an infant-mortality region, which is indicated by a decreasing initial failure rate in Figure 1. The optimum ESS time is t0, since at that point, all the infant-mortality defects have been screened out.
If ESS ends before t0, the product still contains infant-mortality defects which will be found by the user of the product. If ESS ends after t0, useful life is consumed without improving the failure rate.
The failure rate may not be zero even after t0. The failures occurring after t0 are not infant-mortality failures though, and they must be dealt with in ways other than ESS.
Many attempts have been made to prescribe standard ESS processes, but since ESS processes are product-specific, the most effective ones are based on a knowledge of the product, its potential defects and the stresses that cause them.1,2,3,4 An effective ESS process generates valuable data which can be used to improve the product as well as to screen out defects. Unfortunately, when ESS is viewed only as a requirement imposed by the customer or the market, its full benefits are not realized.
The compliance-based approach treats ESS like a cookbook process, in which the product is exposed to a standard set of stresses, at standard levels, for standard lengths of time. Little attention is given to the failure mechanisms, to how they are distributed in time, or to how the failure data can be used to improve the product. Compliance-based ESS provides few benefits other than satisfying a customer-imposed requirement.
Compliance-based ESS users can incur unnecessary expense. Table 1, 2 shows a typical ESS program implemented by a manufacturer of aerospace electronics equipment.
From a physics-of-failure point of view, these conditions are practically identical and, with minor modification, they could all be conducted in a single environmental test chamber. Since the compliance-based approach does not bring this level of understanding to the process, each condition was implemented as stated, and a separate test chamber was required for each one.
The physics-of-failure approach to ESS is based on an understanding of the potential types of latent defects in the product, the failure mechanisms and the stresses that cause them.5,6,7 The ESS conditions are set up to precipitate those defects, and the data is used to determine their causes and distributions.
Failure data is communicated to the appropriate design and manufacturing personnel and used to make changes to improve the product. If it is properly set up and operated, a physics-of-failure ESS process can be extremely cost-effective.
Setting Up the ESS Process
ESS is product unique, since each product has its own set of potential defects and since the applied ESS stresses affect each product differently. Even though the ESS process must be set up individually for each product, there are many common features of both products and stresses which cause many ESS processes to be similar.
The stresses applied in ESS are expected to precipitate manufacturing defects. They are not necessarily those the product will see in service. The two most common ESS stresses for electronic products are temperature cycling and vibration. They may be applied sequentially or simultaneously.
It is critical that electronic equipment be monitored during ESS. This is the only way to detect failures under extreme conditions. More importantly, the stresses used in ESS can induce reversible damage not detected in tests conducted at ambient conditions. This induced damage is itself a latent defect, and the ESS process can actually cause early field failures.
Reducing or Eliminating ESS
Since ESS is an inspection step, it does not add value to the product and should be reduced or eliminated as quickly as possible. This cannot be done without proper justification, which requires relevant data.
ESS must be set up to provide data which can be used to reduce or eliminate it. The following eight steps illustrate what should be done:
1. Collect failure rate data during the ESS process.
Failure data must be collected, not just at the beginning and the end, but during the ESS process. It is not enough to know that failures occurred; their time of occurrence must be recorded. Data from all ESS attempts, whether or not there was a failure, must be collected and recorded.
2. Prepare a plot of failure rate vs time.
This is the type of plot shown in Figure 1. If the failure rate decreases with time, there is an opportunity to reduce it if proper product improvements can be made.
If the curve is constant, or if it increases with time, the ESS process cannot be effective because either there are no infant-mortality defects or the wrong stresses, or levels thereof, are being used. If this is the case, ESS should be modified or discontinued and some other means of product improvement must be implemented.
ESS may be conducted anywhere in the manufacturing process flow. Table 2 shows some examples of the types of stresses used for ESS at the component, subassembly, assembly and system levels for electronic equipment.8 Table 38 shows the types of defects which may be detected by temperature cycling and vibration.
The specific levels of ESS stresses are selected to precipitate the relevant defects in a relatively short time, and yet not consume a significant portion of the life of nondefective items. For electronic equipment, the lower end of the temperature cycling range is usually from -40(degree)C to -50(degree)C, and the upper end is from +75(degree)C to +85(degree)C.
The rate of temperature change can also be important. Figure 2 illustrates the effects of temperature rate-of-change on surface-mount transistor lifting.7 Selecting the vibration level can be quite challenging, especially if the defects are susceptible to a range of frequencies.
In general, multiaxis, repetitive shock vibration is much more effective and efficient than single-axis vibration. Simultaneous temperature cycling and vibration also are much more efficient than either separate or sequential application of the two stresses.
3. Analyze failures and separate them according to failure mechanism.
All failures must be analyzed in order to take corrective action. It is truly amazing that many ESS operations do not include any structured method to analyze the failures and to provide the results to those who can take the proper corrective action.
4. Prepare plots of failure rate vs time for each failure mechanism.
After this is done, the criteria of Step 2 must be applied to each failure mechanism. Again, only failure mechanisms with decreasing failure rates can be attacked with ESS.
5. Improve the product.
Without using the data generated by ESS to improve the product, including design, components, materials and processes, there is no hope of reducing or eliminating the ESS process. If the staff responsible for the ESS process is not the staff responsible for designing and manufacturing the product, it is important that good communication take place between the two groups.
6. Collect and analyze ESS data for the improved product.
If the proper steps have been taken to improve the product, then the area under the infant-mortality region of the failure rate vs time curve should be smaller. This may result from either a reduced slope of the curve or from a shorter time in which it reaches a constant failure rate.
7. Modify ESS conditions to reflect the new failure rates.
As failure mechanisms are eliminated, the stresses that precipitate them may be eliminated. If they occur in shorter times, then the duration of the ESS process may be shortened.
In some cases, additional stresses or increased levels may have to be introduced to detect failure mechanisms which were not expected. If this is the case, care must be taken to avoid introducing irrelevant failures.
8. Reduce or eliminate ESS as warranted.
If the ESS process has been set up properly, and if the proper data is collected and used effectively, it will result in a continuously improving product. Eventually, a point will be reached where the ESS process may be reduced significantly or eliminated entirely. It may also be possible to reduce the frequency of ESS by going from a 100% screen to a sample screen.
The effectiveness of ESS ultimately must be evaluated economically. This analysis is based on the cost to conduct ESS, the cost of field failures, and the frequency of occurrence of field failures.7,9,10,11,12,13,14 ESS costs include the cost of capital equipment, the recurring cost of conducting the process, the cost of analyzing and repairing failures, and the risk of actually introducing new failures into the product. The benefit is in the reduced costs of field failures.
Effectiveness of ESS
The references contain many examples of the successful use of ESS. AT&T called its process environmental stress testing (EST) to emphasize the fact that the company used the results to make product improvements.15 The process combined temperature step stress and temperature cycling between -20(degree)C and +70(degree)C for circuit card assemblies.
Figure 3 shows a plot of failures vs the number of cycles in the EST process. From the data in Figure 3, the investigators concluded that the optimum number of temperature cycles was 16.
In addition to the improvement in outgoing quality, the investigators tracked field failure results. They reported a five-fold improvement in product which had been exposed to EST, compared to product which was not exposed to EST.
Although some ESS practitioners believe that the process should always be conducted on 100% of the product, a sample EST process has been implemented successfully.15
One two-stage ESS process for laser diodes was comprised of a steady-state burn-in at 165(degree)C and 10 kA/cm216 The results showed that unscreened lasers had a medium lifetime of about 600 h, compared to about 6,000 h for screened lasers. for 2 h prior to assembly, and a second steady-state burn-in at 70(degree)C for 150 h after assembly.
In another study on laser diodes, AlGaAs laser diodes were exposed to an ESS process consisting of operation under power in inert atmospheres.17 The results are shown in Table 5. Again, significant improvement in operating reliability was obtained for products which had been exposed to ESS.
If a product has a very low failure rate, the design and operation of the ESS process can be quite complex. McClean reported the use of a technique called highly accelerated stress audit to screen printed-circuit card assemblies.18 The screening stresses were temperature cycling and vibration, with power being applied during the process. As the name implies, the test was applied on a sample basis.
As noted in these examples, the development and operation of an ESS process must be highly customized to the product being screened. Perhaps the greatest benefit of ESS is the hands-on knowledge and experience about the product gained by those who design and manufacture it. For this reason, it is not a good idea to assign the ESS process to a reliability department or a third-party screening organization with limited capability to change the design or manufacturing processes.
Alternatives to ESS
ESS is effective only when the product has an infant-mortality region. If this is not the case, other methods must be used. Some other methods which also involve the application of stresses are ongoing reliability testing (ORT), ongoing accelerated life testing and periodic requalification.
ORT exposes a small sample, for example, less than 1% of production on a regular basis, to stresses at or slightly above the operating range for periods ranging from a few days to a few weeks. All failures are analyzed, and the data is used to improve the product. At the conclusion of the test, the surviving samples are shipped as regular product.
Ongoing accelerated life testing is similar to ORT, except that the stresses are somewhat higher, and the test is continued until the samples fail. Since this is a destructive test, the sample sizes may be somewhat smaller than those of ORT, especially if the product is an expensive one.
Periodic requalification involves the repetition of the qualification procedure, or an abbreviated version of it, on a periodic basis (usually once or twice per year). This type of test had its beginning in some of the U.S. military standards. Since periodic requalification does not involve a wide range of sample lots and since it is expensive, it is losing popularity.
The overall purpose of ESS is to assure that, once a product is qualified, there will be no uncontrolled variations in the individual items during the production phase. The application of stresses is necessary to detect some defects which cannot be observed by functional or visual observation.
The only realistic way to develop and operate an effective ESS process is to use the physics-of-failure approach. This requires an understanding of the product, and knowledge of the types of defects and the types of stresses which precipitate them.
Almost by definition, a significant amount of trial and error is associated with developing efficient ESS processes; but once the basic knowledge is gained, it can be applied to a wide range of products. In most cases where ESS has been implemented, it has proven to be quite effective in reducing overall product costs.
1. MIL-STD-2164 (EC), Military Standard Environmental Stress Screening Process for Electronic Equipment.
2. DoD-HDBK-344 (USAF), Environmental Stress Screening of Electronic Equipment.
3. Environmental Stress Screening Guide, Technical Report No. AD-A206, U.S. Army, Ft. Belvoir, VA, January 1989.
4. Environmental Stress Screening Guidelines for Assemblies, Institute of Environmental Sciences, March 1990.
5. Pecht, M., and Lall, P., "A Physics of Failure Approach to Burn-In," Proceedings of the ASME Winter Annual Meeting, 1993.
6. Lambert, R.G., "Case Histories of Selection Criteria for Random Vibration Screening," The Journal of Environmental Sciences, January-February 1985, pp. 19-24.
7. Smithson, S.A., "Effectiveness and Economics--Yardsticks for ESS Decision," Proceedings of the Institute for Environmental Sciences, 1990.
8. Mandel, C.E.N., Jr., "Environmental Stress Screening," Electronic Materials Handbook, Vol. 1, ASM International, Materials Park, OH, 1989, pp. 875-876.
9. Smith, W.B., and Khory, N., "Does the Burn-In of Integrated Circuits Continue to be a Meaningful Course to Pursue?," Proceedings of the 38th Electronic Components Conference, IEEE, 1988, pp. 507-510.
10. Pantic, D., "Benefits of Integrated-Circuit Burn-In to Obtain High Reliability Parts," IEEE Transactions on Reliability, Vol. R-35, No. 1, 1986, pp. 3-6.
11. Shaw, M., "Recognizing the Optimum Burn-In Period," Quality and Reliability Engineering International, Vol. 3, 1987, pp. 259-263.
12. Huston, H.H., Wood, M.H., and DePalma, V.M., "Burn-In Effectiveness - Theory and Measurement," Proceedings of the International Reliability Physics Symposium, IEEE, 1991, pp. 271-276.
13. Suydo, A., and Sy, S., "Development of a Burn-In Time-Reduction Algorithm Using the Principles of Acceleration Factors," Proceedings of the International Reliability Physics Symposium, IEEE, 1991, pp. 264-270.
14. Trindade, D.C., "Can Burn-In Screen Wearout Mechanisms?: Reliability Modeling of Defective Subpopulations--A Case Study," Proceedings of the International Reliability Physics Symposium,IEEE, 1991, pp. 260-263.
15. Parker, P.T., and Harrison, G.L., "Quality Improvement Using Environmental Stress Screening,"AT&T Technical Journal, July-August, 1992, pp. 10-23.
16. Chik, K.D., and Devenyi, T.F., "The Effects of Screening on the Reliability of AlGaAs/GaAs Semiconductor Lasers," IEEE Transactions on Electron Devices, Vol. 35, No. 7, July 1988, pp. 966-969.
17. Tang, W.C., Altendorf, E.H., Rosen, H.J., Web, D.J., and Vettiger, P., "Lifetime Extension of Uncoated AlGaAs Single Quantum Well Lasers by High-Power Burn-In in Inert Atmospheres,"Electronics Letters, Vol. 30, No. 2, January 20, 1994, pp. 143-145.
18. McClean, H., "Highly Accelerated Stressing of Products With Very Low Failure Rates,"Proceedings of the Institute of Environmental Sciences, 1992.
About the Author
Lloyd Condra wrote this article while employed as a consultant to Hanse Environmental, Inc. Today, he is a Principal Engineer at Boeing Company in Seattle. Previously, he was affiliated with AT&T Bell Labs, Medtronics and Eldec. Mr. Condra is a graduate of Leigh University with an M.S. degree in material engineering, and is the author of two technical reference books. Hanse Environmental, Inc., 235 Hubbard St., Allegan, MI 49010, (269) 673 8638.