INL/CON-10-18643

PREPRINT

Balance Calibration A
Method for Assigning a
Direct-Reading
Uncertainty to an
Electronic Balance
2010 NCSL International Workshop and
Symposium
Michael Stears
July 2010

This is a preprint of a paper intended for publication in a journal or

proceedings. Since changes may be made before publication, this
preprint should not be cited or reproduced without permission of the
author. This document was prepared as an account of work
sponsored by an agency of the United States Government. Neither
the United States Government nor any agency thereof, or any of
their employees, makes any warranty, expressed or implied, or
assumes any legal liability or responsibility for any third partys use,
or the results of such use, of any information, apparatus, product or
process disclosed in this report, or represents that its use by such
third party would not infringe privately owned rights. The views
expressed in this paper are not necessarily those of the United
States Government or the sponsoring agency.

Balance Calibration A Method for Assigning a Direct-Reading Uncertainty

to an Electronic Balance
Speaker/Author: Mike Stears
Idaho National Laboratory
P.O. Box 1625, Mail Stop 4137
Idaho Falls, ID 83415
Phone: 208-526-2343
Fax: 208-526-5462
Abstract
Like many calibration laboratories, our laboratory provides calibrations for a wide range of
instrumentation. For the most part, we calibrate the instrumentation to the manufacturers
published specifications. As one of our services, we provide calibrations of electronic balances
for customers within our company; the calibrations are performed at the customers location. In
our experience, most of our customers are not using their balance as a comparator, but simply
putting an unknown quantity on the balance and reading the displayed value. Manufacturers
specifications for balances typically include specifications such as readability, repeatability,
linearity, and sensitivity temperature drift; but, what does this all mean when the balance user
simply reads the displayed mass value and accepts the reading as the true value? What is the
uncertainty in the measurement? This paper discusses a method for assigning a direct-reading
uncertainty to a balance based upon the observed calibration data and the environment where the
balance is being used.
1.

Introduction

The method for assigning a direct-reading uncertainty to an electronic balance, discussed in this
paper, requires close interaction with the customer regarding the environmental conditions where
the balance is in use. The customers uncertainty requirements for measurements performed
with the balance are discussed and assessed against the operating environment. The amount of
rigor the customer is able to apply to monitoring and controlling the operational environment has
a significant impact on the magnitude of the uncertainty applied to the balance.
Uncertainties of the standard weights, used for field calibration of balances, are determined over
the typical range of environmental conditions encountered. Environmental conditions are
assessed and stabilization time for the standards is determined at the time of calibration.
Measurements are then performed and the resulting data is used, along with other uncertainty
contributions, to arrive at an assigned uncertainty for the balance. EA-10/18 Guidelines on the
calibration of non-automatic weighing instruments [1] is an excellent resource for discussion of
uncertainty contributions and methods to arrive at an expanded uncertainty. A spreadsheet,
designed specifically for this process, is used by our laboratory to simplify the process for the
calibration technician and greatly reduce the chance for an error in the calculation of the assigned
uncertainty. Sections of the spreadsheet will be used throughout this paper to assist in the
explanation of the process. Methods for determining the magnitude of the expanded uncertainty
in this paper are somewhat conservative in order to simplify the process and minimize the cost of
2010 NCSL International Workshop and Symposium

calibration. The methodology used is not intended for calibrations in a laboratory-type

environment requiring the lowest uncertainties attainable.
2. Uncertainty of the standards
Due to the fact that the calibration is not performed in a highly-controlled laboratory
environment, an uncertainty analysis is performed for each standard weight covering the range of
environmental conditions typically encountered in the field. The weight uncertainties and
calibrated values are not modified, based upon observed conditions at the time of calibration, as
long as the observed conditions are within the range identified in the analysis. Ranges of
environmental values typical for our calibrations are (15.6 to 26.7) C, (20 to 60) %RH, and
(83.77 to 85.84) kPa. The uncertainty contributions considered are illustrated in Table 1.
Table 1. Example standard weight uncertainty analysis.
Nominal Mass
(grams)
Assumed Air
Density
3
(g/cm )

Contribution

Weight
Calibration

Drift

200

0.001200

2 sigma
uncertainty
(micrograms)

57.5

42.53

Nominal Density (g/cm )

Deviation from INL

3
nominal (g/cm )

Distribution

Normal

Rectangular

7.95

Property #:

0.000200

Divisor

1.732

Date:

Adjusted
Component

xxxxxx

xx/xx/xxxx

Notes

28.75

From calibration report.

If multiple weights are used, add
the uncertainties together.

24.56

Based on calibration history. If

multiple weights are used, RSS
the drift values.

Convection
( 1 C)

80.000

Rectangular

1.732

46.189

Air Buoyancy

36.7925

Rectangular

1.732

21.2428

Weight Density

Rectangular

1.732

Taken from Table F2.1 in EA10/18. Interpolation required for

values not listed. Use 1 g/g for
weights < 10 g.
Nominal air density at INL  (1.0
3
0.034) mg/cm . Valid for: (83.77
to 85.84) kPa, (15.6 to 26.7) C,
(20 to 60) %RH.

Included in weight calibration

uncertainty.

Combined Uncertainty =

63.36

micrograms

Expanded Uncertainty (k = 2)

126.7

micrograms


The contribution for convection is based on information taken from Table F2.1 in EA-10/18.
The table lists nominal values of mass and the estimated change in apparent mass as a result of
temperature differences between the standard weight and the balance being calibrated. Table
F1.2 in EA-10/18 lists acclimatization times necessary to reach specified levels of temperature
equilibrium. The uncertainty for each weight is calculated with different contributions for
2010 NCSL International Workshop and Symposium

convection based on temperature differences between the weight and the unit under test (UUT).
Different uncertainties are calculated so that only an appropriate amount of equilibration time for
the particular calibration is allowed for efficiency. Our laboratory has made the decision to
calculate the standard weight uncertainties at temperature differentials of 1 C and 5 C. The
manufactures repeatability and linearity specifications are used to estimate a target
uncertainty for the specific balance. The target uncertainty is only used to determine the level
of temperature equilibrium that will be enforced prior to performing the calibration. The target
uncertainty is calculated as: target = [(2 * repeatability) + linearity] / range capacity. As a
general rule, a temperature difference (between the standard weights and the balance) of < 5 C
will be enforced for balances where the target uncertainty is  10 g/g. A temperature difference
of < 1 C will be enforced for balances where the target uncertainty is < 10 g/g. These are
simply guidelines used by our laboratory; they can easily be modified to meet specific needs.
The contribution for air buoyancy is based on the maximum difference between typical air
densities at our location and standard air. Corrections are not made at the time of calibration;
the full uncertainty resulting from the range of air densities is included for each standard weight.
3. Preliminary Operations
Upon arrival at the calibration location, the temperature inside of the case containing the
standard weights is compared to the temperature at the location of the balance. Based upon the
balance capability, a stabilization time is determined as previously discussed. The temperature at
the balance is observed throughout the calibration process.
The calibration environment is discussed with the customer at this time to determine a reasonable
temperature range over which a balance self-calibration is valid. This temperature range is
included in the uncertainty analysis and must be monitored by the balance user when the balance
is in use. The balance user must perform a daily self-calibration prior to use and any time the
temperature changes beyond the limit included in the uncertainty analysis; the change is relative
to the temperature at the time the last self-calibration was performed. Some balances perform a
self-calibration automatically with changes in temperature. If a balance requires an external
weight to perform the self-calibration, a calibrated weight is provided to the customer for this
purpose if not already available. For balances that do not have a self-calibration function, a
multiplier is included in the uncertainty analysis; the calibration interval may be reduced as well
until the data justifies changes to the interval or uncertainty.
If the balance has more than one range, it should be verified with the customer which ranges are
to be calibrated. It should also be discussed with the customer whether the full range of the
balance is used or whether calibration of a limited range is desired; this could improve the
uncertainty assigned to the balance for the limited range. The normal usage of the balance,
regarding weighing schemes, should be discussed to determine whether the standard weights
should be measured as increasing or decreasing loads. Prior to taking data, the balance should be
checked to ensure it is level. The balance should be exercised by placing a weight approximately
equal to the capacity of the balance on the weighing pan at least once. If available, perform a
self-calibration function for the balance.

2010 NCSL International Workshop and Symposium

4. Calibration
Once the details of the calibration have been determined, the measurements are taken. Table 2
illustrates the spreadsheet used for the calibration process.
Table 2. Example calibration spreadsheet.
Temperature Range (user specified):
Balance Range:

500

Resolution (Decimal Points):

Balance Temperature Sensitivity Coefficient:

grams

Assigned weight for performance of self-cal:

digits

g/g

internal

Linearity Test

Repeatability Test

Calibrated conventional mass of weight (grams)

As-found (g)

As-left (g)

Reading Number

Value (grams)

0.0099985

0.0099

0.0099

300.0003

0.0999996

0.0999

0.0999

300.0002

1.0000066

1.0000

1.0000

300.0001

9.9999734

10.0000

10.0000

300.0001

100.0000926

99.9999

99.9999

300.0001

300.0001938

300.0003

300.0003

300.0002

500.0004327

500.0005

500.0005

300.0001

300.0001

300.0003

10

300.0002

Standard Deviation =
Mfg Repeatability spec:

0.0000823

grams

0.00012

grams

Corner Loading

Linearity Result

Quadrant

Reading

Maximum deviation from calibrated weight:

CNTR

300.0001

BL

300.0001

Weight Uncertainty

BR

300.0001

Largest uncertainty of all weights used (k=2):

FL

300.0002

FR

300.0001

Mfg linearity spec (k=2):

RSS Weight Uncertainty & Max Deviation:

0.000192638

grams

0.0002

grams

0.0002568

grams

0.000321023

grams

Greater of standard deviation or mfg repeatability spec:

0.0001200

grams

Greater of (RSS Weight Uncertainty & Max Deviation) or mfg linearity spec:

0.000321023

grams

Enter balance readability(d):

0.0001

grams

0.0005000

grams

Temperature drift of sensitivity:

2

Expanded uncertainty calculated as [2 * SQRT((Repeatability) + (Linearity/1.732) + (Readability/3.46) + (Temperature Drift/1.732) )]

Expanded uncertainty =
Temperature:

23.6

Humidity:

42

Property Number:

Multiplier (enter 1 if self-cal performed):

S&CL Assigned Uncertainty =

Date:

Manufacturer:

Customer:

Model:

Location:

Serial No.:

Comments:

2010 NCSL International Workshop and Symposium

0.000729

grams (k = 2)

If no self-cal being performed prior to use, multiply uncertainty by a minimum of 5.

1
0.00073

grams (k = 2)

The yellow shaded areas of the spreadsheet are the fields required to be filled in by the
calibration technician. The remaining fields are automatically filled in by functions built into the
spreadsheet.
4.1 General Information
There are seven buttons along the bottom of the spreadsheet (not shown), the appropriate button
is selected to set the resolution of the spreadsheet to match that of the balance being calibrated
prior to taking data. The temperature range, over which a self-calibration is valid, is entered as
discussed with the customer. The manufacturers specification for the temperature sensitivity
coefficient is then entered. The combination of these two entries is used in the calculation of the
assigned uncertainty. Next, the assigned weight for the performance of self-calibration is
entered. If internal weights are not used, the asset number of the assigned external weight is
entered. The balance range, or sub-range, being calibrated is also entered. See Table 3.
Table 3. General information.
Temperature Range (user specified):
Balance Range:

500

grams

Balance Temperature Sensitivity Coefficient:

Assigned weight for performance of self-cal:

g/g

internal

4.2 Repeatability
Once the required stabilization time had been allowed, the balance has been exercised, and the
self-calibration has been performed; the repeatability test is started. Repeatability tests are
performed with a test load  50 % of the range being calibrated [1]. Ten measurements are taken
of the test load with a zero-check performed prior to each measurement of the test load. The
weight is placed in the center of the weighing pan. The standard deviation of the measurements
is calculated by the spreadsheet and the manufacturers specification is entered by the calibration
technician as shown in Table 4.
Table 4. Repeatability test.
Repeatability Test
Reading Number

Value (grams)

300.0003

300.0002

300.0001

300.0001

300.0001

300.0002

300.0001

300.0001

300.0003

10

300.0002

Standard Deviation =
Mfg Repeatability spec =

2010 NCSL International Workshop and Symposium

0.0000823

grams

0.00012

grams

4.3 Eccentricity Test (Corner Loading)

Eccentricity tests are performed to provide the information to the customer; however, the results
are not normally included in the assigned uncertainty of the balance. The same test load that was
used for the repeatability test may be used. The load is measured in the center of the pan and
then in each of the four quadrants successively as shown in Figure 1. The balance is zeroed
between readings as necessary. If the manufacturer provides a specification for eccentricity, the
readings are analyzed for conformance to the specification; corrective action can be taken as
necessary after all as-found data for the entire calibration is recorded.
Corner Loading (Eccentricity)
Quadrant

Reading

CNTR
BL

300.0001
300.0001

BR

300.0001

FL

300.0002

FR

300.0001

BR

BL

BL

CNTR
FL

BR

CNTR
FR

FL

FR

Figure 1. Datasheet entry and load positions.

4.4 Linearity Test
The linearity test section of the spreadsheet does not actually verify the manufacturers linearity
specification for the balance; it is simply used to determine the errors of indication for the
balance over the range being calibrated. The measurement data could be used to verify the
linearity specification, but this is not included in the methodology our laboratory typically uses
to assign a direct-reading uncertainty. The test weights are selected to provide nominal
indications of 1 on each consecutive digit of the balance readout starting with the second least
significant digit; the minimum nominal value test weight used is 1 mg. Multiple test points are
normally selected over the most significant decade of the balance range. The test weights are
selected through a pick list for each entry as shown in Table 5.
Table 5. Pick-list selection of test weights.

2010 NCSL International Workshop and Symposium

The calibrated values of the test weights and their associated uncertainties are in a second
worksheet behind the calibration worksheet; the pick lists are populated from the test-weight
worksheet. Having the test weights selected through a pick list greatly reduces the risk of having
an error in the calibrated value of any weight listed as a standard. The uncertainty for each test
weight used in the calibration is automatically pulled into the calibration worksheet for inclusion
in the uncertainty analysis and determination of the assigned direct-read uncertainty. Calibrated
values and uncertainties of the test weights are updated when the weights are calibrated; this is
the only time the values of the test weights are manually entered into the worksheet. The values
and uncertainties entered for the test weights are double checked by a second person at the time
of entry.
Once the test weights have been selected, the weights are measured on the balance and the asfound data is recorded. When all of the data is recorded, it is reviewed to determine whether
adjustment of the balance is necessary. The manufacturers linearity specification is used as a
basis for determination of adjustment; the uncertainty of the test weights should be considered
when making this decision. If the balance is not adjusted, the as-found data is duplicated in the
as-left column of the worksheet. If it is determined that the balance is to be adjusted, the
manufacturers adjustment procedure is followed. After adjustment, the test weights are
measured again and the data recorded in the as-left column of the spreadsheet. If the adjustment
achieved the desired results, all of the data for the calibration is complete.
5. Uncertainty Calculation
The determination of the assigned uncertainty can be handled in different ways. The uncertainty
calculation in the calibration spreadsheet produces a single uncertainty for the entire range
calibrated. The uncertainty determination is based on a combination of the measurement data
and the manufacturers specifications.
The worst-case deviation from the calibrated weight values in the as-left data is root-sumsquared (RSS) with the largest individual uncertainty of the test weights used; the result is
compared to the manufacturers linearity specification and the larger of the two is included as the
linearity contribution in the combined uncertainty. The standard deviation of the repeatability
data is compared to the manufacturers repeatability specification and the larger of the two is
included as the repeatability contribution in the combined uncertainty. These comparisons and
calculations are performed automatically by the spreadsheet.
The balance readability (scale interval or d) is entered by the technician for inclusion in the
combined uncertainty. The spreadsheet multiplies the temperature sensitivity coefficient by the
user-specified temperature range to arrive at the temperature drift of sensitivity contribution to
the combined uncertainty.
Individual uncertainty contributions are converted to a one-sigma level through division based on
the distribution. The repeatability contribution is considered using a normal distribution; the
linearity, readability (d/2), and temperature drift of sensitivity are considered using a rectangular
distribution. The one-sigma uncertainty contributions are combined using the RSS method to
arrive at a combined uncertainty. The combined uncertainty is multiplied by a k-factor of two to
2010 NCSL International Workshop and Symposium

produce the expanded uncertainty. Calculation of the expanded uncertainty is handled

automatically by the spreadsheet. The spreadsheet illustration of the uncertainty contributions is
shown in Table 6.
Table 6. Combining uncertainty contributions.
Linearity Result
Maximum deviation from calibrated weight:
Mfg linearity spec (k=2):

0.000192638

grams

0.0002

grams

0.0002568

grams

Weight Uncertainty
Largest uncertainty of all weights used (k=2):

0.000321023

grams

Greater of standard deviation or mfg repeatability spec:

RSS Weight Uncertainty & Max Deviation:

0.0001200

grams

Greater of (RSS Weight Uncertainty & Max Deviation) or mfg linearity spec:

0.000321023

grams

Enter balance readability(d):

0.0001

grams

Temperature drift of sensitivity:

0.0005000

grams

Expanded uncertainty calculated as [2 * SQRT((Repeatability) + (Linearity/1.732) + (Readability/3.46) + (Temperature Drift/1.732) )]

Expanded uncertainty (k=2):

0.000729

grams

If no self-cal being performed prior to use, multiply uncertainty by a minimum of 5.

Multiplier (enter 1 if self-cal performed):
S&CL Assigned Uncertainty (k = 2):

1
0.00073

grams

The expanded uncertainty is modified by a multiplier if the balance does not have a selfcalibration function. If this is the initial calibration of the balance, in the current environment,
the multiplier is normally set at five. The multiplier can be modified by the calibration
technician based on calibration history of the balance and acceptable risk to the customer. The
multiplier could be significantly reduced if the customer is willing to incorporate a check
standard into the measurement process to monitor performance of the balance. All customers are
encouraged to use check standards to provide measurement assurance regardless of the
capabilities of the balance. If the balance has a self-calibration function, the multiplier is set at
one. The product of the expanded uncertainty and the multiplier is the direct-read uncertainty
assigned to the balance.
If the balance being calibrated has more than one range, the calibration process described is
repeated for each of the ranges. A separate spreadsheet is generated for each range to arrive at
an individual uncertainty assigned to a specific range.
The assigned uncertainty does not include any contribution for samples weighed by the
customer. Any additional uncertainty specific to a sample such as density of the sample, static
electricity, magnetism, moisture, temperature, etc. must be analyzed by the customer for impact
on the measurement results. Calibration personnel should be available to assist with questions
regarding uncertainty contributions.

2010 NCSL International Workshop and Symposium

6. Conclusions
In our experience dealing with customers regarding balance use and calibration, most did not
fully understand all of the uncertainty considerations or how to make a meaningful estimate of
their measurement uncertainty. This process was implemented to engage the customer in the
calibration process and provide a meaningful estimate of the balance uncertainty in the
environment where it is being used. This process has become an educational experience for the
customer as well as our calibration personnel. The process outlined is one method to assign a
direct-reading uncertainty to an electronic balance; additional tests may need to be added to meet
all of the requirements for specific applications. This process is meant to provide an efficient
and consistent method for calibration personnel to perform the calibration and assign the
uncertainty in the field. The uncertainty calculations are meant to cover a wide range of
environmental conditions and are not meant for balance use in a tightly controlled environment
where the lowest uncertainties attainable are desired.
7. References
1.

European co-operation for Accreditation, Guidelines on the calibration of nonautomatic

weighing instruments, EA-10/18, 2005.

2010 NCSL International Workshop and Symposium