How much test is enough?
Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 10/1/2005
One fundamental question always haunts us: When are we finished testing? The answer is simple, at least in theory. We test products until they work.
Unfortunately, reaching that goal proves far more complicated in practice. Last May, a major maker of embedded defibrillators announced that a flaw in some units could cause a small number of them to short-circuit and fail. In the days that followed, other makers admitted similar problems. All of them also announced that they had already changed their manufacturing processes to eliminate the defects.
 |
| Especially in complex or high-reliability products, the cost of test and repair approaches infinity as the desired yield approaches 100%. |
Stig Oresjo, senior test-strategy consultant for Agilent Technologies, agrees with the consequence issue and proposes four other "c's": complexity, confidence, coverage, and cash-flow. The first three define how much test you need. The last two—coverage and cash flow—serve as tradeoff indicators. Higher complexity (more defect opportunities) demands more comprehensive test.
Also, carefully consider your confidence that your test strategy accomplishes all of your goals. Estimating test coverage at any step includes assessing the cost of subsequent failures even before products ship to customers. Oresjo concludes that the dearth of easy answers is the main justification for test engineering.
John VanNewkirk, president and CEO of CheckSum, looks at "fitted" rather than "enough" test. He suggests fitting a test strategy to the product's fault spectrum in addition to its other characteristics and advocates modifying the strategy based on that spectrum. Fitted test implies testing not only enough, but also for the right defects. Knowing the fault spectrum permits the use of layering test and inspection techniques so each can identify the faults that it can find most efficiently. VanNewkirk contends that some manufacturers concentrate too much on complicated "chimera" faults (that rarely occur), ignoring more common faults such as cold solder joints that their tools cannot find.
Yet—as with the defibrillators—no matter how carefully we plan, some faults get through. Preventing that from happening requires constant vigilance and a willingness to change strategies to incorporate new information.
| Reference: |
|
|
Wayne Tustin of the Equipment Reliability Institute has published a new text covering environmental test topics. The 444-page book, Random Vibration & Shock Testing, discusses HALT, HASS, and ESS measurements, and it includes a CD-ROM with video clips and animations. Price: $250. 
