| The design is finished, the manufacturing transition is | | | | addressed. |
| complete, pre-production units have successfully | | | | Test Limits |
| passed acceptance testing, the product is going out | | | | If the equipment is being manufactured to a customer's |
| the door, and the design team is all ready for the next | | | | specification, then those specifications become the |
| project, right? | | | | minimum levels of stress. If the application is critical, |
| Not quite. | | | | then testing above these limits by 10-15 percent is |
| In this age of shrinking product and component size, | | | | recommended. If the equipment is being built to your |
| faster speeds and immense complexities, electronic | | | | own product specifications, then those specifications |
| products are pushing the envelope of reliable | | | | become the test limits. In more complex systems, |
| performance due to, among other things, self-heating | | | | different circuits have different failure rates. These can |
| effects and nanofabrication tolerances. In severe | | | | be estimated by using piece-part reliability ratings from |
| operating environments, reliable performance margins | | | | such references as MIL-HDBK-217. Those areas with |
| can easily be exceeded, resulting in catastrophic | | | | the highest failure rates should be pinpointed for |
| system failures. Therefore, it is important that products | | | | maximum stress. |
| are tested over and above anticipated ambient | | | | Time & Cost |
| conditions to ensure specified performance in real-life | | | | This can vary widely, depending on the extent of |
| environments. | | | | testing. Once the test limits have been defined, the |
| Unfortunately, extra reliability testing, because it occurs | | | | program has to be implemented through hardware, |
| towards the end of a project when little or no funds | | | | software or a combination of both. A typical stress |
| remain to perform adequate stress testing, sometimes | | | | test team would consist of a reliability engineer, a |
| gets orphaned. But ignoring or under-funding reliability | | | | hardware design engineer and a computer |
| stress testing can end up being one of the largest cost | | | | programmer, if required. Once the team has defined |
| factors in a product's life cycle. | | | | the scope of the stress testing, then the time and |
| How Much Testing Is Enough? | | | | material estimates will determine the cost of the |
| So the question becomes, how much testing is needed | | | | program. |
| to confidently project long-term reliability? The scope | | | | A note of caution: The design engineer who conceived |
| of reliability stress testing varies widely from product to | | | | the circuit or system to be tested is NOT the engineer |
| product, depending mostly on the complexity and | | | | who should design the test program. All too often, the |
| criticality of the system. In mission critical applications, | | | | system designer does not want his or her system to |
| such as tactical military situations or life support | | | | fail, so the tendency is to design a test that the system |
| equipment requirements, extensive stress testing | | | | will pass. "Design-to-pass" is poor criteria for devising a |
| should be performed to ensure maximum reliability | | | | reliability test and will undoubtedly end up giving the |
| over the life of the product. Less testing is adequate | | | | user a false sense of security. It is much better to |
| for less critical applications. | | | | follow a "design-to-fail" principle. This will end up yielding |
| Testing microprocessor-based equipment presents its | | | | more realistic results. Unfortunately, design engineers |
| own set of challenges. Some believe that it takes a | | | | find it difficult to design a test circuit that may perhaps |
| smart test program to test a smart system. Not | | | | turn up deficiencies in their designs, so the test |
| necessarily. For example, a systems engineer once | | | | program should be left to independent, impartial test |
| proudly described to me the test program he had | | | | engineers. |
| designed for a sophisticated processor-based system. | | | | Capital Equipment |
| He claimed that he had included a series of neatly | | | | If the plan calls for in-house stress testing, then the |
| partitioned sub-routines that checked out all of the | | | | manufacturer must purchase or rent the required test |
| system's functions using test vectors, flashing | | | | equipment needed to simulate the stress environment. |
| indicators, plus a computer printout of the pass/fail | | | | Again, the actual equipment will be determined by the |
| condition of practically every circuit in the system. He | | | | required environmental limits (e.g., a temperature |
| challenged me to induce a fault in the system that his | | | | chamber for thermal stresses, a vibration table for |
| eloquent test program could not locate. I disabled the | | | | shock and vibration performance, an electromagnetic |
| main processor clock oscillator, which defeated his | | | | radiation source for EMI/RFI susceptibility). In almost all |
| entire program. The point is that elaborate testing is not | | | | cases, subjecting the unit under test to temperature |
| required. The most effective tests are usually the | | | | variations is a basic requirement. Other environmental |
| simplest, straight-forward ones. | | | | test functions are based on the anticipated operating |
| From the Simple to the Complex | | | | conditions of the equipment in its final environment. |
| Stress testing can vary from simply placing | | | | Outside Agencies Can Help |
| components in a cookie tin and subjecting them to a | | | | If buying or leasing capital equipment is not practical, |
| few hot-cold temperature cycles to fully exercising a | | | | environmental testing labs may be an option. Many |
| system in its intended environment over maximum | | | | have fully staffed reliability design engineers willing to |
| specified limits. No matter how simple or how | | | | assist in designing the overall stress test program. Third |
| complicated the equipment or device is, stress testing | | | | party advisors such as Nerac, whose analysts include |
| at some level is a basic requirement. The only question | | | | engineers experienced in the field of reliability stress |
| is how much or how little is required. | | | | testing, can assist with assessing and implementing an |
| A comprehensive stress testing program weighs the | | | | efficient and economical stress test program and can |
| risk factors anticipated at maximum rated operating | | | | help find qualified labs to conduct the tests. |
| conditions and assigns a probability of failure with each. | | | | Designing, implementing and conducting efficient stress |
| The factors with the highest probability of failure should | | | | testing can sometimes be as daunting as the design of |
| be thoroughly tested early in the stress test program. | | | | the system it is testing. It is complex, and many |
| Less stringent testing can be performed in the areas | | | | manufacturers often underestimate or overlook the |
| of low probability of failure. | | | | importance of such a program. This can cost dearly in |
| How does one go about configuring an effective | | | | fixing latent defects or funding product or system |
| stress test program? Three questions should be | | | | recall expenses down the road. |