Fluke 5560A Service Manual page 15

Calibrator
Calibration Verification of the Product
Figure 2. Test Limits Established by Guardbanding
STD
UUT
-SL TL Nom Meas + TL + SL
There are many ways to develop test limits from decision rules. Two of these rules are shown as
examples to develop test limits through guardbanding.
ILAC G8 (ISO 14253-1) Decision Rule
In this guardband strategy example, determine the test limit by the subtracting the uncertainty of the
measurement from the specification.
Test Limit = Specification Limit - Expanded Uncertainty of Measurement
Obtain the specification limits from the performance verification procedure or compute them with
information from Determine Specification Limits for other Calibration Intervals as guidance.
Once the calibration laboratory determines the expanded uncertainty for a measurement, compute the
test limits for this method as follows:
Using the example test point in the previous section, for the 1-year, 99 % confidence specification for
1 mA at 1 kHz, the specification limits are 350 nA. If for example, the expanded uncertainty of
measurement at this test point was 55 nA, the test limit would be:
Test Limit = 350 nA - 55 nA = 295 nA
This creates upper and lower test limits of 0.999705 mA to 1.000295 mA. If the measured value from
the calibration were between these limits, by this decision rule the Product would be in tolerance or
indicated as Pass.
If the measured value obtained was between 295 nA (the test limit) and 350nA (the specification limit)
this is known as an indeterminate measurement by ISO 14253-1. Some organizations elect to call this
a conditional pass as it is more likely that the measurement indicates an in tolerance condition than not.
If the measured value is >350 nA, but <405 nA (the sum of the specification limit plus the uncertainty)
ISO 14243-1 indicates that this is an indeterminate measurement as well. Some organizations choose
to call this a conditional fail because there is still a possibility that the measurement is in tolerance.
Most organizations consider this an out of tolerance result because although the value may exist
anywhere with the interval of the measured value plus and minus the uncertainty, the best estimate of
the value is the measured value.
11