Specifications

135Calibration and Automated Calibration
The goal of calibration is to maintain a
piece of equipment in its most accurate
state. The goal of automated calibration
is to improve eciency and consistency
of the calibration process, while
minimizing the down time required
to verify equipment performance.
Accuracy vs. Precision
The terms accuracy and precision are
often used synonymously, but they are
not the same thing. Both are, however,
needed to achieve the best results. We
can illustrate the dierences between
these two terms through the following
example. To measure the performance
of a particular system, one can plot the
results of a large number of samples over
time on a graph and note the dierences
between the actual results and the
desired result (Figure 1). Accuracy is the
measure of how close the mean of the
total set of results is to the desired result.
Precision is a measure of the spread
of these results relative to this mean.
Precision only addresses how dispersed
the results are, not how far they are
from the average of the desired value.
Calibration
Calibration usually addresses accuracy
and less often precision. From the above
discussion it is evident that calibration
may not have any eect on precision,
because other circuit parameters such
as noise may have an inuence on
precision and no amount of calibration
will reduce the spread of values. This is
of course not always the case, such as
in light beam focusing. When a beam is
correctly focused, its spread is reduced.
This is, of course, not always the case.
For complete basic calibration, it is often
required to correct for both oset and
span (gain). This requires calibration
at more than one point. If a system is
linear, calibration at two points will
suce since two points dene a line
(Figure 2). If a system is nonlinear,
more calibration points are needed.
Calibration is the process of adjusting
circuit parameters, such as osets
and gains, to make equipment meet
specications or a standard. All
organizations producing electronic
goods must either design with
high-precision components or
use some form of calibration.
All electronic products must pass at
least minimal signal testing prior to
shipping to ensure that the product
works out of the box. A rigorous
test and calibration process also
reduces liability from performance
errors and provides a paper trail that
shows that industry and regulatory
requirements have been followed.
Although new products may meet
strenuous requirements for calibration,
due to the eects of use, wear, and
environmental conditions, over
time products may no longer meet
specications. For some products the
eect is easily seen: a cell phone that no
longer receives calls, or a hard drive that
loses data. For others, e.g., a voltmeter
with a small drift, the eect cannot
be easily seen, but the impact may be
costly. Or, in the case of an insulin pump,
the impact may be even dangerous. For
many types of industrial (and medical)
electronic equipment, calibration is
an on-going process and is the reason
why many products are now being
designed with self-calibration circuitry.
For control devices used in a production
environment, a proper calibration
process uses test equipment that has
been certied to standards traceable
to a government agency. In the U.S.
this agency is the National Institute of
Standards and Technology (NIST). This
type of certied calibration requires
the services of a certied metrology
lab. The lab will not only calibrate the
equipment based upon recognized
standards, but will also provide reports
as part of their service. These reports
prove that the equipment has been
measured and adjusted relative to a
chain of standards traceable back to
the government’s master standards.
Figure 1. Accuracy and precision are two very different things.
ACCURACY
PRECISION
MEASURED
VALUES
NUMBER OF
OCCURRENCES
RESPONSE
STIMULUS
GAIN ERROR,
NO OFFSET ERROR
CALIBRATED:
NO GAIN OR
OFFSET ERROR
OFFSET ERROR,
NO GAIN ERROR
Figure 2. For proper basic calibration, the system response to
stimuli must be corrected for both offset and gain errors. Offset
errors do not produce a zero output for a zero input. Gain errors
(when no offset error remains) show more deviation from the
expected results at larger input stimuli.