Campbell Manufacturing CR10 ユーザーズマニュアル

When both junctions of a thermocouple are at the
same temperature, there is no voltage produced
(law of intermediate metals).  A consequence of
this is that a thermocouple cannot have an offset
error; any deviation from a standard (assuming the
wires are each homogeneous and no secondary
junctions exist) is due to a deviation in slope.  In
light of this, the fixed temperature limits of error
(e.g., +1.0

°C for type T as opposed to the slope

error of 0.75% of the temperature) in the table
above are probably greater than one would
experience when considering temperatures in the
environmental range.  In other words, the
reference junction, at 0

°C, is relatively close to the

temperature being measured, so the absolute
error (the product of the temperature difference
and the slope error) should be closer to the
percentage error than the fixed error.  Likewise,
because thermocouple calibration error is a slope
error, accuracy can be increased when the
reference junction temperature is close to the
measurement temperature.  For the same reason,
differential temperature measurements, over a
small temperature gradient, can be extremely
accurate.

In order to quantitatively evaluate thermocouple
error when the reference junction is not fixed at
0

°C, one needs limits of error for the Seebeck

coefficient (slope of thermocouple voltage vs.

temperature curve) for the various
thermocouples.  Lacking this information, a
reasonable approach is to apply the percentage
errors, with perhaps 0.25% added on, to the
difference in temperature being measured by
the thermocouple.

The accuracy of a CR10 voltage measurement
is specified as 0.2% (0.1% 0 to 40

°C) of the full

scale range being used to make the
measurement.  The actual accuracy may be
better than this as it involves a slope error (the
error is proportional to the measurement being
made though limited by the resolution).  The
error in the temperature due to inaccuracy in
the measurement of the thermocouple voltage
is worst at temperature extremes, where a
relatively large scale is necessary to read the
thermocouple output.  For example, assume
type K (chromel-alumel) thermocouples are
used to measure temperatures at 600

°C.  The

TC output is on the order of 24.9 mV, requiring
the +25 mV input range.  The accuracy
specification of 0.1% FSR is 25 uV which is a
temperature error of about 0.60

°C.  In the

environmental temperature range with voltage
measured on an appropriate scale, error in