Page 19 - Troubleshooting Analog Circuits

6 I. First Things First

Conversely, if the op amp’s VouT was a few dozen millivolts in error, you might
start checking the resistors for their tolerances. You might not check for an open-
circuited or wrong-value capacitor, unless you checked the circuit’s output with a
scope and discovered it oscillating!! So, in any circuit, you must study the data-
your “clues”-until they lead you to the final test that reveals the true cause of your
problem.
Thus, you should always first formulate a hypothesis and then invent a reasonable
test or series of tests, the answers to which will help narrow down the possibilities of
what is bad, and may in fact support your hypothesis. These tests should be perform-
able. But you may define a test and then discover it is not performable or would be
much too difficult to perform. Then I often think, “Well, if I could do that test, the
answer would either come up ‘good’ or ‘bad.’ OK, so I can’t easily run the test. But if
I assume that I’d get one or the other of the answers, what would I do next to nail
down the solution? Can I skip to the next test??”
For example, if I had to probe the first layer of metal on an IC with two layers of
metal (because I had neglected to bring an important node up from the first metal to
the second metal), I might do several other tests instead. I would do the other tests
hoping that maybe I wouldn’t have to do that probing, which is rather awkward even
if I can “borrow” a laser to cut through all the layers of oxide. If I’m lucky, I may
never have to go back and do that “very difficult or nearly impossible” test.
Of course, sometimes the actual result of a test is some completely unbelievable
answer, nothing like the answers I expected. Then I have to reconsider-where were
my assumptions wrong? Where was my thinking erronegus? Or, did I take my
measurements correctly? Is my technician’s data really valid? That’s why trouble-
shooting is such a challenging business-almost never boring.
On the other hand, it would be foolish for you to plan everything and test nothing.
Because if you did that, you would surely plan some procedures that a quick test
would show are unnecessary. That’s what they call “paralysis by analysis.” All things
being equal, I would expect the planning and testing to require equal time. If the tests
are very complicated and expensive, then the planning should be appropriately com-
prehensive. If the tests are simple, as in the case of the 128 resistors in series, you
could make them up as you go along. For example, the list above of resistors #80,
112, 120, 104,88, or 72 are nominally binary choices. You don’t have to go to ex-
actly those places-an approximate binary search would be just fine.

You Can Make Murphy’s Law Work for You
Murphy’s Law is quite likely to attack even our best designs: “If anything can go
wrong, it will.” But, I can make Murphy’s Law workfor me. For example, according
to this interpretation of Murphy’s Law, if I drive around with a fire extinguisher, if I
am prepared to put out any fire-will that make sure that I never have a fire in my
car? When you first hear it, the idea sounds dumb. But, if I’m the kind of meticulous
person who carries a fire extinguisher, I may also be neat and refuse to do the dumb
things that permit fires to start.
Similarly, when designing a circuit I leave extra safety margins in areas where I
cannot surely predict how the circuit will perform. When I design a breadboard, I
often tell the technician, “Leave 20% extra space for this section because I’m not sure
that it will work without modifications. And, please leave extra space around this
resistor and this capacitor because I might have to change those values.” When I
design an IC, I leave little pads of metal at strategic points on the chip’s surface, so
that I can probe the critical nodes as easily as possible. To facilitate probing when