You forgot to read the data sheet. Application
engineers routinely shout "RTFDS" as they hang up the telephone after
dealing with a customer's enquiry. Extracting the information which is
implicit, rather than just the explicit details, is a skill well worth
You forgot Ohm's Law. The resistance of a PC
track is not zero, and sometimes the difference matters. Conversely,
if you are measuring low currents, the leakage resistance of
insulators is also important.
You forgot to provide a path for the bias
current. Sometimes your greasy fingerprints will provide a path in the
prototype, leading to surprises in the clean(er) final version.
You forgot to check the stray resistances,
inductances, and capacitances of the final (crowded) PCB, and assumed
that if it worked on the breadboard it would work anywhere.
You forgot that EMI and RFI occur everywhere
and did not filter your supplies and input/output leads.
You forgot to consider the effects of
temperature variations on circuit elements, including the effects of
differing temperature coefficients in nominally identical components.
You forgot to verify that the circuit would
tolerate having its supplies (and signals) applied in any order (and
with any value of dv/dt). Or you forgot to ensure that it would not be
exposed to power/signal sequences and rates that it could not
tolerate. You may even have forgotten that switching power supplies
are not necessarily as noise free as a battery.
You forgot that circuits do not necessarily
contain power-up reset, but may require additional circuitry to ensure
that they stabilize in the correct mode when they are turned on.
You forgot that circuits do not turn on
instantly: Capacitors must charge, and it is often necessary for the
circuit to stabilize thermally, especially in precision circuits.
You forgot that not all circuits are stable
when driving a reactive load. An output stage that will drive a wide
range of resistive loads may oscillate if it sees capacitance. Your
circuit may work when its load is close by and fail when it must drive
a few meters of cable.
You forgot that noise, like death and taxes, is
universal. Every ADC has quantization noise and every resistor has
Johnson noise. This is intrinsic, not the result of imperfections.
You forgot that IC designers are not
necessarily user friendly. EOC does not necessarily imply DRDY, and
don't dare to assume that CS only selects the chip - it may also reset
something on its trailing edge. Again and again I say, "RTFDS!"