Dc Accuracy Test Considerations - HP 3455A Operating And Service Manual

Section [V
+ (0.003% + 0.001%) = + 0.004% of reading
Similarly, at one tenth of full scale (0.1 V) the least signifi-
cant digit (10 microvolt) is equal to 0.01% of reading so the
accuracy specification is:
+ (0.003% + 0.01%) = + 0.013% of reading
These specifications do not include the temperature coeffi-
cient that must be added if the instrument is operated out-
side of the 22°C to 24% C range.
4-17. DC ACCURACY TEST CONSIDERATIONS.
4-18. Because of the high de accuracy of the 3455A, a
precision dc calibration standard is required to verify that
it meets its de accuracy specifications. To thoroughly test
the performance on all ranges, the standard must be capable
of delivering outputs within the range of 0.10000 V de to
1000.000 V dc. The accuracy of the standard must be such
that its errors do not introduce significant uncertainties in
the 3455A test readings. Ideally, the accuracy of the stand-
ard should be ten times better than the 3455A specifica-
tions being tested — a ten to one error reduction nearly
eliminates measurement uncertainties caused by the stand-
ard. To test accuracy specifications on the order of
+ 0.005% of reading, however, a standard with a specified
accuracy of + 0.0005% (5 ppm) would be required. Since
this type of accuracy, over the range needed to completely
test the accuracy of the 3455A, is generally not available
outside of a standards laboratory, some compromises may
be required. If you have access to primary in-house (NBS
certified) standards or have calibrated transfer standards
that are capable of delivering the required output voltages,
we recommend that you use them. If you do not have
access to such facilities you may, depending on your specif-
ic accuracy requirements, choose to do one of the follow-
ing:
a. Use a de calibration standard that is four or five
times more accurate than the 3455A specifications to
be tested. (A discussion of the potential uncertainties is
given in following paragraphs.)
b. Use a highly stable calibrated standard and add the
correction factors (given on the calibration chart) to the
3455A test readings.
c. Send the 3455A to an -hp- Service Center or some
other NBS-certified standards facility for calibration.
4-19. Several of today's commercially available de catibra-
tion standards provide the output voltage range and resolu-
tion needed to test the performance of the 3455A but they
are not, in general, an order of magnitude more accurate
than the 3455A. When using such standards it is important
to be aware of the uncertainties or "ambiguities" that may
be encountered. These potential ambiguities are described
in the following paragraphs.
4-20. First, consider the case where a digital voltmeter
(DVM) is to be tested for a full-scale accuracy of + 0.01%
42
Model 3455A
of reading on its 1-volt range. The DVM is connected to a
de calibration standard whose specified accuracy is +
0.001% of setting and with the standard set to + 1.00000 V,
the DVM reads +0.99992 V which is 0.008% low. The dc
standard's specified accuracy is ten times better than the
specification being tested and at 1 V its maximum error
contribution to the DVM reading is 10 microvolt or
0.001%. If the standard is 0.001% low the actual DVM
error is - 0.007%; if it is 0.001% high, the actual DVM error
is - 0.009%. In either case the DVM is within its specifica-
tion and, since this measurement is not a calibration but is
only intended to verify that the DVM meets its specifica-
tion, the standard's error can be ignored.
4-21, But what if the DVM reading is + 0.999908 V? Here,
the DVM appears to be in tolerance (0.0092% low) but the
margin is only 0.0008% which is less than the 0.001% maxi-
mum allowable error contribution of the standard. If the
standard's output is 0.001% low, the actual DVM error is
- 0.0082% rather than - 0.0092% so the DVM is within its
specification. If, on the other hand, the standard's output is
0.001% high, the actual DVM error is - 0.102% and the
DVM is slightly out of tolerance. Chances are good that the
DVM is within its specification but the only way to tell for
sure is to use a more accurate standard. As the example
points out, there are regions of ambiguity even when the
standard is ten times more accurate than the instrument
being tested. With a ten-to-one error reduction, however,
these regions are relatively narrow. In this case, the DVM
could be out of tolerance but if so, its maximum out-of-
tolerance error is only - 0.0002%. As long as the DVM
reading is within specified tolerances, the maximum DVM
error that can exist is + 0.011% which is the sum of the
maximum DVM error and the maximum allowable error
of the standard. A potential deviation of + 0.001% from the
DVM specifications could, in many cases, be acceptable.
Also, if the standard has been recently calibrated and is
known to be well within its specification, readings in the
narrow ambiguous regions may reflect marginal DVM per-
formance or indicate the need for adjustment.
4-22. Now suppose the de standard's specified accuracy is
+ 0.0025% — only four times better than the + 0.01% DVM
accuracy specification. If the DVM reading is + 0.999890
volt, it appears that the DVM is 0.011% low. However, if
the de standard is 0.002% low (well within its specification)
the DVM is only 0.009% low and is in tolerance. Con-
versely, if the DVM reading is + 1.00081 V the DVM
appears to be 0.0081% high and well within its specifica-
tion. But if the standard is 0.0023% low, the actual DVM
error is + 0.014% and the DVM is out of tolerance.
4-23. Figure 4-1 shows how the error tolerances of the
standard combine with those of the DVM to produce the
positive and negative ambiguous regions used in the pre-
ceding examples. From Figure 4-1, the following observa-
tions can be made:
a. If the DVM reading is in tolerance by a percentage
that is greater than the maximum allowable error of the
standard, the DVM is definitely within its specification.