Analog Angle Article

When testing is too easy, it’s easy to test too haphazardly

Everyone is talking about the costs of modern medicine, from many different perspectives and with different objectives. One of the things that strikes me about today's medical advances is that one of the reasons for its high cost is that we have done too good a job of improving non-invasive, no- or low-risk tests. Many of our medical “test and measurement” procedures make it fairly easy to say “let's do that test”. This is a worthy result at the individual level, but perhaps not as good in the aggregate.

This was especially apparent when I overheard a group of senior citizens talking at breakfast at a nearby table (absolutely unintentional, I assure you) and they were discussing their various aches and pains. It seemed like the answer to almost any discomfort was “so I'll go for a scan, what can it hurt?” And technically speaking, they were right, since an MRI or CAT scan entails only a minor discomfort with no surgery, no recovery time, and no apparent risk. The result is a lot of tests are done almost on a whim, when maybe the test is unneeded or excessive. It's the old story: in an ironic way, the system has become a victim of its own success and advances.

This incident got me thinking about test and measurement of our electronic products, especially loaded printed circuit boards (PCBs)–or printed wiring boards (PWBs), as the official language folks prefer we refer to them. Several decades ago, many of the ICs on these boards were not soldered on directly, but instead mounted via sockets. Today's tight board spacing, fine-pitch ICs, and cost pressures have eliminated the use of sockets in production runs, except for some specialized on-board components.

This was both good and bad in terms of test. In the prototype stage, sockets made it easy to swap out a suspect part (it was always a vendor's defective part, since we never, ever fried one!) or substitute an alternate-source device. It also made it easier to add small fixes, jury-rigged add-ons, and specialized test-probe gimmicks.

But it also encouraged a certain kind of lazy thinking as the product moved to production, Since boards could be tested by non-destructive component substitution, unlike removing/replacing today's high-pin-count tight-pitch ICs), the tendency was to not think too much about a problem, just do a lot of component swapping until the problem went away. As a result, the performance verification was not as rigorous as it could have been.

As long as final test found a combination of pats that worked on the PCB, it was OK to ship. Never mind that sometimes the problem was that timing margins were too tight, or error tolerances had built up, hey, the swapping yielded a combination that apparently worked, so why worry?

Of course, once in the field, the unavoidable effects of component aging, drift, temperature, vibration, and the usual factors pushed the PCB into the intermittent or total-failure zone, and the test/substitution cycle started again. And many returns from the field worked fine when they got back to the factory, of course, since so many of them were due to the product being right on the edge of “OK.”

So, what's the lesson for us? It some ways, it is like the medical situation, but in other ways, it is very different. For electronics, it's this: when test is too easy and leads to apparent solutions and success, we can be lulled into thinking too little about the underlying issues. And we can go through the motions of test and evaluation, substituting visible action for up-front thought and assessment.♦

0 comments on “When testing is too easy, it’s easy to test too haphazardly

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.