TECHNICAL PAPERS

North American One Gigaohm

Interlaboratory Comparison:
2006–2008
Jay Klevens
Abstract: Increasingly accurate high resistance measurements are required in electronic measurements. In order to
verify these measurements, an interlaboratory comparison (ILC) was conducted at the 1 gigaohm (109 ohms) resist-
ance level. Participants in the U. S. and Canada included manufacturers, commercial calibration laboratories, and gov-
ernment laboratories. Six different measurement methods were used; all six methods were validated by the ILC. The
National Institute of Standards and Technology performed the opening and closing measurements on the artifacts.
NCSL International provided generous support of this ILC. This paper presents the results, the methods, and the par-
ticipant’s associated uncertainties. Some recommendations are offered for developing more consistent uncertainty deter-
minations for various measurement techniques and for developing a better understanding of leakage currents at high
resistance measurements and guarding against them.

1. Introduction Using measurement results obtained by a national metrology

A good method of assuring a measurement is to compare results institute (NMI) insures that the measurements are traceable to
with another laboratory. A single comparison with another lab- fundamental SI units.
oratory will give an idea if one or the other laboratory is getting Commercial proficiency test providers are excellent resources
valid results, but there is a possibility that both labs are report- for many measurement parameters, but for unusual or little-
ing similar errors. Comparing measurements with a number of used measurements, there are often no commercial proficiency
other laboratories improves the probability of identifying errors. tests available. This was the case at the 1 gigaohm (1 × 109 Ω)
resistance level. Thus an interlaboratory comparison (ILC) was
initiated, with the support of NCSL International, to survey and
Jay Klevens to evaluate the state of measurement at this level in North
Ohm-Labs, Inc. America.
611 E. Carson St. Not that long ago, measurements at the 1 gigaohm level were
Pittsburg, PA 15203-1021 USA a rough, pass-fail sort of test, used in insulation and dielectric
Email: [email protected] proving. Measurements with an uncertainty of 1 % were close

74 | MEASURE www.ncsli.org
 TECHNICAL PAPERS

to practicing state of the art. But the need to measure very small charter and proposal (see appendix for charter and list of par-
currents in ever smaller electronics, along with increasingly ticipants).
accurate high voltage systems, created a necessity for improved The charter defined the ILC’s scope and goal, provided refer-
accuracies at high resistance. After several decades of develop- ences, explained the sponsorship and expenses, and outlined the
ment, high resistance measurements have approached parts-per- overall ILC structure. (See Appendix.) To protect the confiden-
million uncertainties. tiality of the participants’ results, the charter referenced ‘Level
II’ confidentiality, as defined in NCSLI’s Recommended Practice
2. Proposal (RP-15, Sections 4.3 and 4.4). In essence, the participants’
The ILC was designed following NCSLI’s Recommended Prac- results would remain anonymous. Each participant laboratory
tice RP-15 [1], titled, “Recommended Practice for Interlabora- would know its own result, but not that of the other laborato-
tory Comparisons.” A comparable publication dealing with ries. To provide anonymity, all laboratories (except for NIST)
interlaboratory comparisons is ISO 17043, “Conformity assess- were assigned letter codes.
ment — General requirements for proficiency testing,” [2]
which may be purchased from the International Standards 4. Artifacts
Organization (ISO). Much the same guidance for structuring an The pivot laboratory provided two commercially available
ILC is freely available on the National Institute of Standards and 1 gigaohm resistance standards. The measurement connections
Technology (NIST) website. [3] were via BPO (British Post Office) coaxial panel plugs mounted
A draft proposal was developed that outlined the intent and in fluoropolymer (PTFE) discs. As many laboratories did not
structure of the ILC, the region (U. S. and Canada), the time- have mating BPO jacks, the artifacts included accessory adap-
frame (2006-2007), and the nature and purpose (assisting par- tors to BNC male and female connectors.
ticipants in measurement at the 1 GΩ level). The author The standards contained a thermistor in close proximity to
volunteered to serve as the pivot laboratory and to provide the the internal resistor shell. The thermistor had a nominal value
resistors that were used as artifacts. of 10 kΩ at 25 °C. Participants were instructed to measure the
It is important to anticipate and, if possible, to plan in thermistor at the time of test, in order to provide an indication
advance for the behavior of the artifacts over the projected of the temperature inside the resistance standard’s case.
course of the ILC. This is because an ILC may take many The two artifacts, identified with serial numbers 6074 and
months to complete, and because the artifacts will be repeatedly 6075, were packed in a clamshell type foam filled carrying case,
shipped to various laboratories, as well as to the pivot labora- which was then surrounded by cushioning material and packed
tory. The typical problems encountered with resistors are drift in a larger cardboard box for transit.
over time, changes caused by physical shocks, or damage from
misuse. 5. Instructions
Because the artifacts used in the ILC were newly manufac- Once the charter and proposal had been finalized, the partici-
tured resistors, it was important to keep close watch on their pant list was closed. A draft of the ILC instructions and the pro-
performance. Therefore, to monitor drift over the duration of posed measurement worksheet were circulated to the
the ILC, a ‘modified petal’ ILC structure was chosen. In this participants for review and comment. After incorporating some
case, the artifacts were returned to the pivot laboratory for helpful comments, the ILC instructions and worksheet were
checks after every two or three participant’s measurements. finalized, and no subsequent changes were made during the
While this ILC structure would lengthen the duration of the ILC.
ILC, it was hoped that it would hold the artifacts under tighter In the instructions, participants were asked to take less than
measurement control. two weeks to complete the measurements and documentation.
Each participant paid the cost of shipping the artifacts to the However, most participants took longer than the requested time,
next laboratory; otherwise the coordinator paid the costs of the which resulted in the ILC extending for approximately twice its
ILC. NIST was chosen as the reference laboratory to provide originally planned duration. This was the fault of the coordina-
resistance measurements at the beginning and end of the ILC tor, who failed to adequately supervise the schedule. It is impor-
(the opening and closing measurements). NCSLI was solicited tant for an ILC coordinator to press the participants to keep an
for sponsorship and also to assist with the cost of the NIST ILC on schedule.
measurements. The instructions included a checklist for receiving inspection.
To be able to participate, the pivot laboratory and coordina- If everything arrived in good condition, the artifacts were to be
tor would operate ‘blind,’ without access to the NIST measure- placed in a 23 °C laboratory environment. Participants noted
ment data until the conclusion of the ILC. the hours elapsed for temperature stabilization and the temper-
ature of the artifacts at the time of test.
3. Charter Participants were cautioned to use a guarded measurement
Participants for the ILC were solicited at the 2006 Measurement system and to connect the case terminal of the artifacts to
Science Conference (MSC) and at the 2006 NCSLI Conference. ground. The thermistor resistance was noted at the time of
The ILC charter and proposal were circulated to the various measurement, as well as the power to the thermistor.
participants, with a request for comments or suggestions. These Participants were asked to measure each resistor at 10 volts
were gratefully received and were incorporated into the final and at 100 volts. Not all participants were able to measure at

Vol. 4 No. 1 • March 2009 MEASURE | 75

TECHNICAL PAPERS

6074 10 V Linear Values 6074 100 V Linear Values

Date Lab (Ω) 10 V ( Ω ) (Ω) 100 V ( Ω )

9/12/2006 NIST ––– 1 000 001 150. ––– 1 000 000 900.

10/23/2006 Pivot 1 000 000 890. 1 000 001 640. 1 000 003 120. 1 000 001 281.

2/22/2007 Pivot 999 999 610. 1 000 003 098. 999 997 110. 1 000 002 413.

5/20/2007 Pivot 999 998 410. 1 000 004 137. 999 992 050. 1 000 003 221.

10/3/2007 Pivot 1 000 001 450. 1 000 005 762. 1 000 001 500. 1 000 004 484.

2/1/2008 Pivot 1 000 008 110. 1 000 007 208. 1 000 010 340. 1 000 005 607.

3/4/2008 Pivot 1 000 011 520. 1 000 007 590. 1 000 011 400. 1 000 005 904.

6/30/2008 NIST ––– 1 000 009 000. ––– 1 000 007 000.

Table 1. Reference and uncorrected pivot laboratory measurements for artifact 6074.

Artifact 6074
1,000,015,000

1,000,010,000
Resistance, Ω

1,000,005,000

1,000,000,000

10 V
999,995,000
100 V
Weighted
999,990,000
Sep–06 Dec–06 Mar–07 Jun–07 Sep–07 Dec–07 Mar–08 Jun–08

Figure 1. Reference and uncorrected pivot laboratory measurements for artifact 6074 showing the linear interpolation of the NIST data, as
well as the weighted reference values.

both voltage levels. Participants were asked to provide their from a linear interpolation between the opening and closing
uncertainty budgets; this information was used to assist in NIST measurements; they are provided for comparison purposes.
working towards some uniformity in the uncertainty compo- Comparing these two data sets shows that the artifacts and
nents unique to high resistance measurement. All but one par- the pivot lab measurement system remained in control through
ticipant filled out the uncertainty sections of the ILC worksheet. the ILC. No widely outlying measurements appeared. The trend
Several participants submitted extensive uncertainty analyses. of the measured values appears to roughly agree with the linear
The coordinator requested electronic transmittal of all docu- values between opening and closing NIST measurements.
ments for the ILC. Electronic copies of the participants’ docu- The data show that the resistance of each artifact varied over the
ments have been archived by the coordinator, should the 22 month time period time of the ILC. For artifact 6074, based on
participants wish to see a copy of their originals. NIST’s measurements, the resistance increased about 7 µΩ/Ω. For
artifact 6075, the resistance increased about 21 µΩ/Ω. One
6. Results approach to determining reference values for each laboratory’s
After all participants had reviewed their data, and after NIST measurements is to linearly interpolate between the opening and
had completed the closing measurements, the coordinator closing NIST measurements. However, the drift of one artifact did
reviewed and analyzed the measurement results. not appear to be linear, as evidenced from the pivot laboratory’s
Tables and Figs. 1 and 2 show the NIST opening and closing data. Therefore, instead of a linear interpolation, the coordinator
data, together with the uncorrected pivot laboratory measure- chose to weight one artifact’s drift based on the date and value of
ments. These illustrate the drift of the artifacts over the course the intermediate pivot laboratory measurements.
of the ILC. The linear values shown in the tables were calculated Artifact 6074 appears to have dropped by a few µΩ/Ω

76 | MEASURE www.ncsli.org
 TECHNICAL PAPERS

6075 10 V Linear Values 6075 100 V Linear Values

Date Lab (Ω) 10 V ( Ω ) (Ω) 100 V ( Ω )

9/12/2006 NIST ––– 999 999 600. ––– 999 999 200.

10/23/2006 Pivot 999 996 000. 1 000 000 935. 1 000 000 140. 1 000 000 498.

2/22/2007 Pivot 1 000 012 470. 1 000 004 909. 1 000 006 440. 1 000 004 360.

5/21/2007 Pivot 1 000 009 310. 1 000 007 776. 1 000 008 060. 1 000 007 146.

10/2/2007 Pivot 1 000 011 690. 1 000 012 140. 1 000 0116700. 1 000 011 389.

2/2/2008 Pivot 1 000 020 600. 1 000 016 147. 1 000 017 740. 1 000 015 283.

3/5/2008 Pivot 1 000 017 770. 1 000 017 189. 1 000 022 570. 1 000 016 296.

6/30/2008 NIST ––– 1 000 021 000. ––– 1 000 020 000.

Table 2. Reference and uncorrected pivot laboratory measurements for artifact 6075.

Artifact 6075
1,000,025,000

1,000,020,000
Resistance, Ω

1,000,015,000

1,000,010,000

1,000,005,000
10 V
1,000,000,000
100 V
999,995,000
Linear
999,990,000

Sep–06 Dec–06 Mar–07 Jun–07 Sep–07 Dec–07 Mar–08 Jun–08

Figure 2. Reference and uncorrected pivot laboratory measurements for artifact 6075, showing the linear interpolation of the NIST data.

through the first half of the ILC, then to have drifted upwards Where there are no results above a laboratory code, the par-
during the second half of the ILC. Based on these data, a refer- ticipant did not report a measurement at that point. One partic-
ence value for this artifact was derived by assigning a fixed ipant reported excessive noise with one artifact at one
resistance value through the first half of the ILC, and linearly measurement voltage, and this measurement was suppressed
interpolating the rise, fit using the opening and closing NIST from the report. The varying horizontal spacing between meas-
measurements, through the second half of the ILC. urement points indicates the elapsed days between participant’s
Artifact 6075 appears to have drifted upwards roughly lin- measurements. For clarity, pivot laboratory measurements are
early through the ILC. Therefore, the reference value of this arti- not shown in these figures.
fact can be defined as a linear interpolation between the opening From these figures, it is clear that nearly all participants meas-
and closing NIST measurements. These two weighted sets of ured the artifacts well within their claimed uncertainties.
resistance values will be used as the laboratory’s reference However, there were a couple of outlying measurements.
values, against which participant’s results can be compared. Overall, the participants should be pleased with their measure-
These weighted reference values and the participants’ results ment proficiency and with the general state of measurement
are shown in Figs. 3 to 6 and Tables 3 to 6. For artifact number competency at the 1 gigaohm level.
6074, using the above weighting as a zero reference line, the
participant’s measured deviation, in µΩ/Ω, are shown in Figs. 3 7. Proficiency Evaluation
and 4. For artifact number 6075, the weighted reference values A quantitative evaluation of measurement proficiency can be
were again used and results are shown in Figs. 5 and 6. Also obtained by comparing the difference with the reference value
shown in the figures are the uncertainties estimated by partici- to the combined uncertainty of the reference value and the
pants as error bars for k = 2. The first and last points are the uncertainty of each participant laboratory value. [2] The result-
opening and closing NIST measurements. Note that at both ing number, En, is obtained by dividing the difference between
10 V and 100 V, most of the participant’s results agreed with the a participant measurement, x, and the reference value, X, by the
reference value to within about ±20 µΩ/Ω. square root of the sum of the participant laboratory uncertainty,

Vol. 4 No. 1 • March 2009 MEASURE | 77

TECHNICAL PAPERS

Artifact 6074 at 10 V
100
80
∆R, µΩ/Ω, Normalized to

60
Weighted Reference

40
20
0
–20
–40
–60
–80
–100
NIST A B C D E F G H I J K L M NIST

Laboratory

Figure 3. Variation in the participant’s measurements of artifact 6074 at 10 V, normalized to weighted reference values listed in Table 3.

Uncertainty Weighted Deviation

6074 10 V
Date Lab ( µ Ω/Ω ) Reference From Ref. En
(Ω)
k=2 (Ω) (Ω)

9/12/2006 NIST 1 000 001 150. 6.8 1 000 001 150. 0. 0 00

11/8/2006 A 999 995 000. 24.0 1 000 001 150. –6 150. –0.25

12/12/2006 B 999 997 501. 25.0 1 000 001 150. –3 650. –0.14

1/30/2007 C 999 978 010. 33.5 1 000 001 150. –23 140. –0.68

3/27/2007 D 999 932 000. 22.0 1 000 001 150. –69 150. –3.07

4/10/2007 E ––– ––– 1 000 001 150. ––– –––

4/28/2007 F 1 000 002 166. 92.0 1 000 001 150. 1 016. 0.01

8/6/2007 G 1 000 064 483. 35.9 1 000 002 130. 62 350. 1.72

9/8/2007 H 999 981 000. 67.9 1 000 003 110. –22 110. –0.32

9/27/2007 I 1 000 004 000. 15.0 1 000 004 090. – 90. –0.01

11/20/2007 J 1 000 004 000. 25.9 1 000 005 080. –1 080. –0.04

12/19/2007 K 1 000 003 300. 15.0 1 000 006 060. –2 760. –0.17

2/1/2008 L 1 000 008 110. 15.0 1 000 000 040. 1 070. 0.07

2/21/2008 M 1 000 028 000. 601.4 1 000 008 020. 19 980. 0.03

6/30/20008 NIST 1 000 009 000. 6.8 1 000 009 000. 0. 0.00

Table 3. Measurement results for artifact 6074 at 10 V, showing weighted reference values along with En values. Values in red are greater or
less than ±1.0.

Ulab, and reference laboratory uncertainty, Uref, squared. The +1.0 or less than −1.0 means that the participant laboratory
equation is: value lies outside the claimed measurement uncertainty. An En
x−X . of less than ±0.5 shows a comfortable level of measurement
En = (1)
U2lab + U2ref uncertainty for the participant laboratory. En values of each par-
ticipant are shown in Figs. 7 and 8.
An En within the range ±1.0 means that the participant labo-
ratory value agrees with the reference value, within the bounds 8. Temperature Effects
of the claimed measurement uncertainties. An En of greater than The precious-metal oxide resistors used in high resistance stan-

78 | MEASURE www.ncsli.org
 TECHNICAL PAPERS

Artifact 6074 at 100 V

50

40

30
∆R, µΩ/Ω, Normalized to
Weighted Reference

20

10

–10

–20

–30

–40

–50
NIST A B C D E F G H I J K L M NIST

Laboratory

Figure 4. Variation in the participant’s measurements of artifact 6074 at 100 V, normalized to weighted reference values listed in Table 4.

Uncertainty Weighted Deviation

6074 100 V
Date Lab ( µ Ω/Ω ) Reference From Ref. En
(Ω)
k=2 (Ω) (Ω)

9/12/2006 NIST 1 000 000 900. 6.8 1 000 000 900. 0. 0 00

11/8/2006 A 1 000 000 000. 24.0 1 000 000 900. –900. –0.04

12/12/2006 B ––– ––– 1 000 000 900. ––– –––

1/30/2007 C ––– ––– 1 000 000 900. ––– –––

3/27/2007 D 999 952 000. 22.0 1 000 000 900. –48 900. –2.17

4/10/2007 E 999 995 610. 39.0 1 000 000 900. –5 290. –0.13

4/28/2007 F ––– ––– 1 000 000 900. ––– –––

8/6/2007 G 999 996 092. 35.9 1 000 001 660. –5 570. –0.15

9/8/2007 H 1 000 009 000. 67.9 1 000 002 430. 6 580. 0.10

9/27/2007 I 1 000 003 000. 15.0 1 000 003 190. –190. –0.01

11/20/2007 J 1 000 006 000. 25.9 1 000 003 950. 2 050. 0.08

12/19/2007 K 1 000 001 900. 15.0 1 000 004 710. –2 810. –0.18

2/1/2008 L 1 000 010 340. 15.0 1 000 005 480. 4 870. 0.31

2/21/2008 M 1 000 028 000. 601.4 1 000 006 240. 21 760. 0.04

6/30/20008 NIST 1 000 007 000. 6.8 1 000 007 000. 0. 0.00

Table 4. Measurement results for artifact 6074 at 100 V, showing weighted reference values along with En values. Values in red are greater
or less than ±1.0.

dards tend to have higher temperature coefficients of resistance time of test. The artifacts in this ILC had a temperature coeffi-
(TCR) than most wire wound standard resistors. Due to the cient of resistance of about +20 µΩ/Ω / °C. (Note that newer
higher TCR for these standards, significant errors might be standards have lower TCR values than this value.)
introduced by variation in the temperature of the artifact at the Comparing the reported resistance with the reported tempera-

Vol. 4 No. 1 • March 2009 MEASURE | 79

TECHNICAL PAPERS

Artifact 6075 at 10 V
100

80
∆R, µΩ/Ω, Normalized to

60
Weighted Reference

40

20

–20

–40

–60

–80

–100
NIST A B C D E F G H I J K L M NIST

Laboratory

Figure 5. Variation in the participant’s measurements of artifact number 6075 at 10 V, normalized to weighted reference values listed in
Table 5.

Uncertainty Weighted Deviation

6075 10 V
Date Lab ( µ Ω/Ω ) Reference From Ref. En
(Ω)
k=2 (Ω) (Ω)

9/12/2006 NIST 999 999 600. 6.8 999 999 600. 0. 0 00

11/8/2006 A 999 998 000. 24.0 1 000 001 460. –3 460. –0.14

12/12/2006 B 999 990 405. 25.0 1 000 002 560. –12 160. –0.48

1/30/2007 C 999 988 570. 33.5 1 000 004 160. –15 590. –0.46

3/27/2007 D 999 926 000. 22.0 1 000 005 980. –79 980. –3.55

4/10/2007 E ––– ––– 1 000 006 440. ––– –––

4/28/2007 F 999 939 547. 92.0 1 000 007 120. –67 580. –0.73

8/6/2007 G 1 000 067 545. 35.9 1 000 010 350. 57 200. 1.58

9/8/2007 H 999 996 000. 67.9 1 000 011 330. –15 300. –0.23

9/27/2007 I 1 000 014 000. 15.0 1 000 011 980. 2 018. 0.13

11/20/2007 J 1 000 013 000. 25.9 1 000 013 510. –510. –0.02

12/19/2007 K 1 000 015 900. 15.0 1 000 014 680. 1 200. 0.08

2/1/2008 L 1 000 020 600. 15.0 1 000 016 150. 4 450. 0.28

2/21/2008 M 1 000 038 600. 601.4 1 000 016 800. 21 800. 0.04

6/30/20008 NIST 1 000 021 000. 6.8 1 000 021 000. 0. 0.00

Table 5. Measurement results for artifact 6075 at 10 V, showing weighted reference values along with En values. Values in red are outside
the range ±1.0.

ture might show the temperature dependence of the artifact. the predicted value. The temperature distribution versus resist-
However, there does not appear to be a correlation shown by the ance deviation appears to be random. Figures 9 and 10 superim-
data, as seen in Figs. 9 and 10. One laboratory reported a temper- pose the weighted reference resistance to 23 °C in order to show
ature as low as 20.2 °C but the reported value was quite close to the relative measured deviation and temperature differences.

80 | MEASURE www.ncsli.org
 TECHNICAL PAPERS

Artifact 6075 at 100 V

50

40

30
∆R, µΩ/Ω, Normalized to

20
Weighted Reference

10

–10

–20

–30

–40

–50
NIST A B C D E F G H I J K L M NIST

Laboratory

Figure 6. Variation in the participant’s measurements of artifact number 6075 at 100 V, normalized to weighted reference values listed in
Table 6.

Uncertainty Weighted Deviation

6075 100 V
Date Lab ( µ Ω/Ω ) Reference From Ref. En
(Ω)
k=2 (Ω) (Ω)

9/12/2006 NIST 999 999 200. 6.8 999 999 200. 0. 0 00

11/8/2006 A 1 000 011 000. 24.0 1 000 001 000. 10 000. 0.41

12/12/2006 B ––– ––– 1 000 002 080. ––– –––

1/30/2007 C ––– ––– 1 000 003 630. ––– –––

3/27/2007 D 999 959 000. 22.0 1 000 005 410. –46 410. –2.06

4/10/2007 E 1 000 002 370. 39.0 1 000 005 850. –3 480. –0.09

4/28/2007 F ––– ––– 1 000 006 510. ––– –––

8/6/2007 G 1 000 002 922. 35.9 1 000 009 650. –6 728. –0.19

9/8/2007 H 1 000 023 000. 67.9 1 000 010 600. 12 400. 0.18

9/27/2007 I 1 000 015 000. 15.0 1 000 011 230. 3 770. 0.24

11/20/2007 J 1 000 023 000. 25.9 1 000 012 720. 10 280. 0.39

12/19/2007 K Excessive Noise N/A 1 000 013 860. ––– –––

2/1/2008 L 1 000 017 740. 15.0 1 000 015 280. 2 460. 0.16

2/21/2008 M 1 000 043 100. 601.4 1 000 015 920. 27 180. 0.05

6/30/20008 NIST 1 000 020 000. 6.8 1 000 020 000. 0. 0.00

Table 6. Measurement results for artifact 6075 at 100 V, showing weighted reference values along with En values. Values in red are greater
or less than ±1.0.

9. Measurement Methods and Uncertainty Cutkosky divider bridge (Measurements International MI

Components 6000); (2 & 3) commercially available teraohmmeter (Guild-
It is interesting to note that the participants used six different line) using both the direct measurement method and the substi-
measurement methods, and that all six methods yielded valid tution method; (4) active arm (or dual source) bridge; (5)
results. The various methods were: (1) commercially available unbalanced Wheatstone bridge; and (6) substitution method

Vol. 4 No. 1 • March 2009 MEASURE | 81

TECHNICAL PAPERS

Artifact 6074
1.5

1.0
Participant En Values

0.5

0.0

–0.5
10 V
–1.0 100 V

–1.5
A B C D E F G H I J K L M

Laboratory

Figure 7. Participant En values when measuring artifact 6074 at both 10 V and 100 V.

Artifact 6075
1.5

1.0
Participant En Values

0.5

0.0

–0.5

10 V
–1.0
100 V

–1.5
A B C D E F G H I J K L M

Laboratory

Figure 8. Participant En when measuring artifact 6075 at both 10 V and 100 V.

using a long-scale meter.1 These measurement methods are tor in series and applies positive voltage to one and negative to
listed in Table 7. the other, proportioned such that roughly zero volts is present
A Cutkosky divider is an R-R2 circuit suited to binary voltage at the junction of the two resistors. The ratio of the positive and
division [4]; in a bridge it is used as an adjustable voltage negative voltages is proportional to the ratio of the two resist-
divider to null against the measurement nodes of a standard and ances. A teraohmmeter measures how long a capacitor requires
unknown resistor connected in series as a fixed voltage divider. to discharge through a resistor. The unbalanced Wheatstone
An active arm bridge places a standard and an unknown resis- bridge was specially built by one participant; its results were
well within its claimed uncertainty. The meter was a long-scale
digital multimeter. These methods all proved satisfactory for
1 Certain commercial equipment, instruments, or materials are identi-
measuring at the 1 GΩ level.
fied in this paper in order to adequately describe the experimental Although there was general consistency in the measurement
procedure. Such identification does not imply recommendation or
results, the measurement uncertainties show less consistency.
endorsement by the author or NCSL International, nor does it imply
that the materials or equipment identified are the only or best avail- Twelve of the thirteen participants submitted uncertainty
able for the purpose. budgets, and all twelve budgets were different. The similarities
and differences therefore warrant discussion.

82 | MEASURE www.ncsli.org
 TECHNICAL PAPERS

Temperature and Deviation, Artifact 6074

26.0 30

25.0 20
Temperature, ºC

24.0 10

23.0 0

22.0 Temperature –10

Deviation
21.0 –20
R Deviation

20.0 –30
A B C D E F G H I J K L M

Laboratory

Figure 9. Temperature and resistance deviation from weighted reference value for artifact 6074.

Temperature and Deviation, Artifact 6075

26.0 30

25.0 20
Temperature, ºC

24.0 10

23.0 0

22.0 Temperature –10

Deviation
21.0 –20
R Deviation
20.0 –30
A B C D E F G H I J K L M

Laboratory

Figure 10. Temperature and resistance deviation from weighted reference value for artifact 6075.

NIST used an active arm bridge; the uncertainty analysis for uncertainties. The reported uncertainty of the pivot laboratory
this method is described in NIST Technical Note 1458, “NIST was 15 µΩ/Ω at k = 2.
Measurement Service for DC Standard Resistors.” [5] A majority of the participants (eight) used a Measurements
NIST transfers the intrinsic standard of a quantum Hall resis- International 6000 bridge. There were notable variations in
tor (around 12.906 kΩ) using a cryogenic current comparator their uncertainty budgets. These uncertainty components are
to the 1 megohm level. From there, 100:1 Hamon-type transfer tabulated in Table 9 for comparison. Note that the reported
standards build up to the 1012 ohms level. NIST’s uncertainty Bridge uncertainty varied from 1 µΩ/Ω to 20 µΩ/Ω. All but one
components, including Type A and Type B components at 1 GΩ, laboratory included the standard deviation of the measure-
using an active arm bridge, are listed in Table 8. Note that the ments. One did not included the uncertainty of the Bridge and
two major sources of uncertainty are the working standard and the uncertainty of the Working Standard used. Only four
temperature (2 µΩ/Ω) each. The overall expanded uncertainty included the temperature uncertainty of the 1 GΩ resistors; only
at k = 2 for NIST was 6.8 µΩ/Ω. three included a leakage current uncertainty.
One participant with an active arm bridge did not provide These eight participants reported final combined (root sum
uncertainty components. The pivot laboratory used an active square) measurement uncertainties (k = 2) ranging between
arm bridge and followed the scheme of NIST’s uncertainty 15 µΩ/Ω and 67 µΩ/Ω. Given the similarity of their systems and
budget, but with higher uncertainty components, mostly for the methods, one would expect this range to be lower. Some recom-
values of the Reference Standard and the Working Standard mendations are discussed in the next section.

Vol. 4 No. 1 • March 2009 MEASURE | 83

TECHNICAL PAPERS

Date Lab Measurement Method a significant source of error. Parasitic current paths act as shunt
resistors, giving falsely low results; this error increases as the
9/12/2006 NIST Active Arm Bridge measurement voltage and resistance increases. To reduce
leakage currents, low voltage measurement connections should
11/8/2006 A MI 6000B be shielded at ground potential; higher voltage measurement
cables should be physically separated from ground. In addition,
12/12/2006 B MI 6000A the shield of the higher voltage measurement cable should be
driven at the measurement voltage. This voltage acts as an elec-
1/30/2007 C MI 6000A
trostatic barrier (a guard) against leakage paths; it should
extend to (and into) the resistor being measured. High measure-
3/27/2007 D MI 6000
ment guards should never be grounded. Low measurement
4/10/2007 E
Fluke 8508A Meter - guards and the cases of both the standard and unit under test
Substitution Method
should be grounded to reduce noise.
4/28/2007 F MI 6000A Laboratories using a teraohmmeter or a long-scale DMM
should include as a Type B uncertainty component either the
8/6/2007 G MI 6000B manufacturer’s specification for resistance measurements or a
characterized uncertainty provided by a calibration report. For
Wheatstone Bridge (10 V)
9/8/2007 H meters used in the direct-measurement method, the measure-
MI 6000B (100 V)
ment uncertainty budget is usually fairly simple: the Type A
9/27/2007 I Active Arm Bridge
component is the standard deviation of a series of measure-
Teraohmmeter – ments; the Type B component is the manufacturer’s accuracy
11/20/2007 J
Substitution Method specification for the measurement device.
12/19/2007 K MI 6000 The substitution method first measures a calibrated working
standard with a meter, then a unit under test, and finally the
2/1/2008 L Active Arm Bridge working standard a second time in order to verify repeatability
and sensitivity. The difference between the working standard
Teraohmmeter –
2/21/2008 M
Direct Measure and the unit under test is thus measured with lower uncertainty
than a direct measurement of the unit under test using the
6/30/20008 NIST Active Arm Bridge
meter. In this method, Type B components include the transfer
Table 7. Measurement methods used by the participant laborato-
uncertainty of the meter and the working standard.
ries and NIST. All participants using these methods properly assigned their
uncertainty components and all achieved En values of less than
10. Recommendations for Uncertainty Budget and 0.5. Based on these results, if less than state-of-the-art uncer-
Measurement Practice tainties are adequate to meet a laboratory’s needs, direct or sub-
It might be helpful to review the major significant uncertainty stitution methods using a meter appear to be valid at the 1 GΩ
components associated with high resistance measurements. level.
Type A components are random, occurring only at the time of To summarize: high resistance measurement uncertainty
test. The standard deviation of a series of measurements is a should include Type A components of measurement repeatabil-
Type A component. If there is measurement history on a unit ity. Type B components should include the working standard
under test, and if there is regression analysis on its drift, the pro- (Rs), bridge ratio, temperature of the unit under test, and
jected drift of the artifact over the calibration interval can also leakage current. The working standard uncertainty should
be considered a Type A component. Generally, these are the only include these components as they accrue from lower resistance
Type A components. levels. The Type A and Type B components should be combined
Type B components are built in to the measurement system. in a root sum square (RSS). This sum should be doubled (k = 2)
For any method using comparison to a standard, the uncertainty to provide approximately 95 % statistical coverage of the result.
of the working standard, Rs, must be included. The uncertainty Leakage in a high resistance measurement system may be
due to temperature of the working standard should also be evaluated by disconnecting the unit under test and taking a
listed, either separately or included as part of the uncertainty measurement of ‘open’ unit-under-test terminals with a
assigned to the standard. The temperature uncertainty is how teraohmmeter or high resistance isolation tester. All terminals
much Rs might vary based on its temperature coefficients and on and connections should be kept clean. High measurement
the temperature stability of its environment during a measure- guards should be driven at measurement potential. High resist-
ment. For an active arm bridge, the NIST uncertainty compo- ance measurements are susceptible to interference from electri-
nents listed above may be used. For a high resistance bridge, the cal noise caused by equipment switching or personnel walking
manufacturer’s specification for the bridge ratio uncertainty in the area. More stable measurements are achieved in areas and
should be listed as a Type B component. For the MI 6000 at times of lower activity. Observing these basic precautions will
bridge, this is 5 µΩ/Ω at 1 GΩ. improve high resistance measurement confidence.
At resistance values above 1 megohm, leakage currents can be

84 | MEASURE www.ncsli.org
 TECHNICAL PAPERS

Type A uncertainty; Assumed to be about 1 µΩ/Ω.

Type B uncertainty at 109 Ohms, in µΩ/Ω (RSS total includes type A value of 1)

Reference Working Bridge Bridge System RSS Total

Detector Temperature Leakage
Standard Standard Ratio Stability Repeatability A+B

0.50 2.0 1.0 0.50 0.10 2.0 0.10 1.0 3.4

Table 8. NIST uncertainty components for the measurement of 1 GΩ standards.

Reported uncertainty components at 1 GΩ

Ω, using a MI 6000 bridge

Lab 1 Lab 2 Lab 3 Lab 4 Lab 5 Lab 6 Lab 7 Lab 8

Type A
Std Dev Std Dev
(None) Std Dev Std Dev Std Dev ––––––––––––––––– ––––––––––––––––– Std Dev Std Dev
Working Standard Stability

Type B

Bridge Bridge Bridge Bridge Bridge Bridge Bridge

(none)
10 µΩ/Ω 1 µΩ/Ω 20 µΩ/Ω 2.5 µΩ/Ω 2.9 µΩ/Ω 5 µΩ/Ω 2.5 µΩ/Ω

Working Working Working Working Working Working Working

Standard Standard Standard Standard Standard Standard Standard

Temp Temp Temp Temp

Leakage Leakage Leakage

Leakage2

Temp 2

Resolution
Working
Standard
Drift
Buildup

Table 9. Reported uncertainty components for eight different laboratories using the same measurement method.

11. Conclusions [3] See NIST Weights and Measures web site:
The 1 GΩ North American ILC surveyed thirteen laboratories http://ts.nist.gov/WeightsAndMeasures/Metrology/roundrobins.cfm
and evaluated six different measurement methods. Most partic- [4] R.D. Cutkosky, “A New Switching Technique for Binary Resistive
ipants’ results agreed with the reference value and were below Dividers,” IEEE Trans. Instrum. Meas., vol IM-27, no. 4, pp.
one half of their claimed uncertainties. Therefore, the overall 421–422, 1978.
state of precision high resistance measurements at 1 GΩ appears [5] R.E. Elmquist et al., “NIST Measurement Service for DC Standard
to be good. Resistors,” NIST Technical Note 1458, National Institute of Stan-
The six measurement methods all appear to be suited to valid dards and Technology, December 2003.
measurement at the 1 GΩ level. Understanding and guarding
against leakage, and applying uniform uncertainty budgets, 13. Appendix
appear to be areas for improvement.
The coordinator would like to thank and congratulate all the 13.1 Final Charter
participants and wishes to extend particular thanks to NCSLI 1. Scope & Goal: This charter is for an inter-laboratory com-
for generous sponsorship of this ILC. parison of two one gigaohm standards. The ILC is designed
to assist in providing each participant confidence in their
12. References measurement at the one gigaohm (1 × 109 Ω) level.
[1] “Recommended Practice for Interlaboratory Comparisons,” 2. References: This ILC is structured according to NCSL RP-
NCSLI Recommended Practice RP-15, NCSL International, 15, “Guide for Interlaboratory Comparisons,” March
Boulder, CO 80301. 1999.
[2] “Conformity assessment — General requirements for proficiency 3. Sponsorship & Expenses: NCSLI will sponsor this ILC by
testing,” ISO 17043, International Standards Organization, contributing towards the cost of NIST’s initial and closing
Geneva, Switzerland. measurements. Any balance of the NIST measurement cost

Vol. 4 No. 1 • March 2009 MEASURE | 85

TECHNICAL PAPERS

will be borne by the coordinator. Each participant will be 13.2 List of Participants
responsible for the cost of shipping the artifacts to the next
laboratory. Air shipment is recommended. Organization Contact Telephone City State Zip
4. Structure: A modified petal structure will be used. NIST Dean Jarrett 301-975-4240 Gaithersburg MD 20899
5. Confidentiality: Confidentiality will be governed by NCSL
Boeing Phil Johnson 206-662-4262 Seattle WA 98108
RP-15, sec. 4.3 & 4.4. The ILC will operate at Level 2 (see
Fluke 972-406-1000
RP-15 section 4.3). Participants will be assigned anony- Electronics
Gary Mote
x766
Carrollton TX 75006
mous lab codes (except NIST, which will be identified as Dave Deaver;
Fluke Corp. 425-466-6434 Everett WA 98203
the reference lab). Participants are expected to maintain the Jorge Martins
613-283-3000
confidentiality of all information and data. Guildline Mike McCain Smiths Falls ON K7A 4S9
x119
6. Publication and Distribution: A final report (subject to the GCS Mike Frisz 407-333-3327 LakeMary FL 32746
above confidentiality restrictions) will be reviewed by all
Keithley Helga Alexander 440-498-3056 Cleveland OH 44139
participants and then submitted to NCSLI for publication
Measurements 613-925-5934
in the journal Measure. Each participant will receive a copy Duane Brown Prescott ON K0E 1T0
International x108
of the final report.
Northrop Redondo
7. Reference Laboratory: NIST will perform initial and Grumman
Bill Cross 310-812-7439
Beach
CA 90278

closing measurements at two voltage levels.

Ohm-Lab Jay Klevens 412-431-0640 Pittsburgh PA 15203
8. Pivot Laboratory: The ILC coordinator will perform inter-
Process
mediate checks. Instruments
Karl Klevens 412-431-4600 Pittsburgh PA 15203
9. ILC Coordinator: Jay Klevens, Ohm-Labs, Inc.
Sandia National Jim Novak;
10. Participants: The ILC participants are laboratories with Laboratories Harold Parks
505-284-3957 Albuquerque NM 87185

low measurement uncertainties at the 1 GΩ level; participa- Redstone

US Army TMDE Vernon Love 256-876-5364 AL 35898
tion is limited to approximately a dozen laboratories to Arsenal

insure a reasonable time limit to the ILC. Wyle Labs Paul Reese 321-494-7907 Patrick AFB FL 32925

86 | MEASURE www.ncsli.org