# MSL Technical Guide 12 Assuring the Quality of Weighing Results

```MSL Technical Guide 12
Assuring the Quality of
Weighing Results
Introduction
The weighing results from electronic balances and
other weighing devices are often of critical importance.
For example they may determine the acceptability of a
product or the outcome of a test. Hence it is important
to have procedures for assuring the quality of these
weighing results.
In this technical guide we discuss the key factors
that affect weighing accuracy and we present inservice checks that can be used to monitor the performance of a balance. While this guide focuses on electronic balances, the principles also apply to other
weighing devices.
This guide assumes that you already have an appropriate balance, set up on a suitably sturdy weighing
bench in a suitable clean environment at a steady
temperature. See [1] for more detail on factors that
can degrade the weighing performance of a balance.
Balance Characteristics
Electronic balances can be remarkably precise.
The resolution of relatively inexpensive laboratory balances can be less than one-millionth of the maximum
error by up to about 0.1 % without careful attention to
factors such as scale adjustment, non-linearity, pan
position error and repeatability.
The scale of a balance itself is usually not perfectly
linear. For example, the balance may read correctly at
Also, the reading of an electronic balance may vary
with the position of the weight on the pan. Reference
[2] explains how to estimate the uncertainty due to pan
position error. In most cases, this uncertainty can be
made negligible by ensuring that the object being
weighed is sufficiently centred on the pan.
Electronic balances measure force. This means
that the scale of an electronic balance is sensitive to
the local value for gravity, which can vary by 0.1 %
depending on where you are in New Zealand [2].
The scale of an electronic balance is also sensitive
to variations in temperature, with a sensitivity typically
in the range from 1 ppm/C to 5 ppm/C (ppm is parts
per million). For example, if a balance with a sensitivity
drift of 4 ppm/C experiences a temperature change of
5 C since the last scale factor adjustment, the balance
reading will change by 20 ppm. That is 2 mg or
0.002 % at 100 g.
For most electronic balances, the scale can be adjusted to compensate for local gravity, for drift in the
electronics, and for variations in temperature. Most
balances now have an automatic scale adjustment
feature, called Cal-Mode or Cal function or equivalent.
When this feature is activated, the balance uses an in-
built weight (or weights) to adjust the scale of the balance. Some balances require an external weight to be
loaded by the operator. The balance scale factor is
adjusted so that the balance correctly measures the
3
mass of objects with a density of about 8000 kg/m .
Regular scale factor adjustment is important for assuring the quality of weighing results. As a guide, it is
good practice to adjust the scale factor each day that
the balance is used. You can assess how frequently
the scale factor needs adjusting by periodically checking the balance accuracy (see below) without adjusting
the scale factor. Some balances automatically readjust the scale factor if the balance temperature
changes significantly.
Balance Calibration
The first requirement in assuring the quality of
weighing results is regular calibration of your balance.
Normally this calibration will be performed by a balance
calibration agent who is accredited by International
Accreditation New Zealand (or equivalent) to ISO/IEC
17025. Prior to calibration, the agent will usually service the balance and set it up to comply with the manufacturer’s specifications at your location. This setup
During a balance calibration, the parameters
measured include the repeatability near full capacity,
linearity (or scale) errors, and pan position errors. The
calibration certificate for the balance gives values for
the repeatability, the maximum pan position error, linearity errors (or corrections to the balance reading),
uncertainty in the linearity errors, and the best accuracy (or limit of performance) of the balance [2].
The best accuracy and linearity corrections are
usually presented in the calibration certificate as a
function of load as shown in Figure 1.
Capacity:
Resolution:
200 g
0.0001 g
Repeatability:
Standard Deviation of 10 repeat readings at 200 g: 0.000 12 g
Linearity & Best Accuracy of Balance:
Nominal
Correction
Expanded
Uncertainty
Best
Accuracy
50 g
0.000 03 g
0.000 22 g
0.000 25 g
100 g
0.000 04 g
0.000 25 g
0.000 29 g
150 g
0.000 27 g
0.000 31 g
0.000 58 g
200 g
0.000 28 g
0.000 40 g
0.000 68 g
Pan Position Error
Measured using a 50 g mass, 25 mm from the centre of the pan.
Figure 1: Part of a balance calibration certificate
Measurement Standards Laboratory of New Zealand  Fax 64 (0)4 931 3002  email: msl@irl.cri.nz  http://msl.irl.cri.nz/
Page 1
Best accuracy includes resolution, repeatability,
measured linearity error and uncertainties in the
standard weights used in the calibration. You can use
the best accuracy values to check that the balance
the balance, you can use the best accuracy as the
expanded uncertainty [3] in the weighing result. The
best accuracy value may be interpolated from the results in the calibration certificate. If you require linearity corrections and they are not given in the calibration
certificate, then contact your calibration agent who will
be able to provide them.
In-Service Checks
The second requirement in assuring the quality of
weighing results is in-service checks between calibrations. These checks are used to confirm that the balance is performing with the required accuracy and to
identify any degradation in performance that might
warrant action (such as servicing and re-calibration).
The history of in-service checks can also be used to
determine the re-calibration interval (up to a recommended maximum of three years).
As a minimum, we recommend:
 a repeatability check every six months at or near
and
 an accuracy check every month at or near full load
(or at several loads over the commonly used range,
if the balance is used to its best accuracy).
The reference values for these checks are usually established (or re-confirmed) directly after the balance
has been calibrated.
Both the repeatability and accuracy checks are performed using calibrated standard weights (see the section on check weights below). This is necessary to
ensure that any changes in repeatability and accuracy
that are apparent from the in-service checks are due to
the balance and not to the check weights.
Other in-service checks may be necessary in some
cases. For example, you may need to periodically
check the pan position error if you are weighing samples that are hard to centre on the pan.
1. Repeatability check. Repeatability is a measure
of the random variations in a balance reading. The
standard uncertainty uR due to repeatability is normally
evaluated as the sample standard deviation of the balance readings for n successive loadings of the same
weight. That is
uR 
n
  r  r   n  1
i 1
th
2
i
(1)
where ri is the i balance reading, and r is the average of the n balance readings. The Excel function
STDEV can be used to calculate uR.
Normally 10 balance readings are recorded (that is,
example, if you normally tare the balance before each
2. Accuracy check. The accuracy check should be
performed following any normal setup procedure for
the balance. For example, if you normally adjust the
scale factor before using the balance, then do this before the accuracy check. Remember that if you use
then you must check the scale factor (and adjust it if
necessary) before each weighing, or batch of weighings.
The accuracy check normally consists of recording
a single reading Q with a calibrated check weight, using your normal weighing procedure.
Acceptance Criteria
Acceptance criteria are defined here for each inservice check, along with the action that will be taken if
these criteria are not met. The acceptance criteria are
normally based on the performance of the balance
directly after servicing and calibration (essentially on
the manufacturer’s specifications).
1. Repeatability criterion. A reference value for the
repeatability uR(ref) is measured directly after a calibration of the balance. For uR(ref), use the greater of the
measured repeatability and 0.41a (the standard uncertainty due to the balance resolution a, see 6.1 in [2]).
Each subsequent in-service check measurement of
repeatability uR(new) is then compared with uR(ref). Here
we use an F-test [4] which gives the acceptance criterion as
uR(new) 
F
uR(ref).
(2)
For example, when each repeatability value is determined from 10 readings and the F-test is based on a
5% probability, the criterion becomes
uR(new)  1.8 uR(ref).
(3)
That is, the in-service repeatability check value must
be no more than 1.8 times the reference value. If criterion (3) is met, you can be reasonably confident that
the value of uR(new) does not indicate any deterioration
in the performance of the balance.
If criterion (3) is not met, then there may be some
deterioration in balance performance. In this case, remeasure the repeatability because there is a 5 %
chance that this criterion will not be met when the balance is performing normally (and vice versa). If the
new repeatability value also doesn’t meet the criterion,
then it is highly probable that there is a problem. An
environmental factor (such as temperature changes,
vibration, or drafts) may have degraded the repeatability or the balance may need servicing.
2. Accuracy criterion. A reference value Qref for
the accuracy check is determined directly after a calibration of the balance by taking the average of q readings. Each subsequent in-service accuracy check value Q (a single reading) is then compared with Qref.
The acceptance criterion is a t-test [4], that is
Measurement Standards Laboratory of New Zealand  Fax 64 (0)4 931 3002  email: msl@irl.cri.nz  http://msl.irl.cri.nz/
Page 2
(4)
For example, for q = 10 and a t-test based on a 5%
probability, this criterion is
0.4
(5)
3. Less stringent criteria. Less stringent criteria
can be used when you are not using the balance to its
best accuracy.
One option is to require more certainty that there
has been a change in the balance performance. This
will increase the acceptance criteria values. For example, you may choose to set the criteria so that the
chance of a failed test when the balance performance
has not degraded is smaller than 5 %. For the repeatability criterion, the constant in (3) is 1.8, 2.3 and 3.2
for 5 %, 1 % and 0.1 % F-tests respectively. For the
accuracy criterion with q = 10, the constant in (5) is
2.4, 3.4 and 5.0 for 5 %, 1 % and 0.1 % t-tests respectively. See reference [4] for other values of F and t.
Another option is to set the acceptance criteria according to the accuracy that you require. Suppose that
you must meet regulatory requirements which specify
an expanded uncertainty of 0.1 %. For example, this
corresponds to an expanded uncertainty of 0.01 g for
the 10 g samples that you weigh. The balance available for the measurement has a best accuracy at 10 g
of 0.0002 g. The purpose of the accuracy check in this
case is to provide assurance that the balance is still
able to meet the regulatory requirement. Here, you
could simply allow a safety factor of three and use
|Q  Qref|  0.003 g.
(6)
as the acceptance criterion for an accuracy check at
10 g. You will need a repeatability check to go with
this accuracy check. Since the purpose of the repeatability check is to identify any degradation in the balance performance, the repeatability check should follow criterion (2) but perhaps with an F-test based on a
0.1 % probability.
Control Charts
It is a good idea to record the results of each inservice check on a control chart, along with the acceptance limit or limits. A control chart provides a
good visual indication of balance performance and can
reveal any shifts or trends in the data. An example is
0.3
0.2
0.1
27-Feb-05
23-Nov-04
19-Aug-04
15-May-04
9-Feb-04
0.0
5-Nov-03
That is, the in-service accuracy check value must differ
from the reference value by no more than 2.4 times the
reference value for repeatability at that load.
If criterion (5) is not met, it is likely that something
is wrong. If a repeat check does not meet the criterion,
then start looking for a problem. Confirm that the balance was properly set up, that you are using the correct check weight, that the accuracy check was performed using the right procedure, that the environment
has not changed and that the balance repeatability is
still acceptable. At this point it may be necessary to
have the balance serviced.
1-Aug-03
|Q  Qref|  2.4 uR(ref).
given in Figure 2 which shows a control chart for balance repeatability checks. In this case the repeatability did exceed the acceptance limit. This was confirmed by two further measurements of the repeatability. After servicing and re-calibration, the repeatability
returned to normal.
Repeatability /mg
1
|Q  Qref|  t 1  uR(ref).
q
Figure 2: Control chart for balance repeatability. The
acceptance limit is shown by the dashed line.
Check Weights
As mentioned, the weights used for the in-service
checks must be suitable for the purpose. The most
common specification for standard weights is OIML
R111-1 [5], which describes the characteristics that
must be met for weights in each of nine different accuracy classes. These characteristics include; density,
shape, construction, magnetism, material and surface
finish. Standard weights that comply with the requirements of OIML R111-1 are available commercially.
Each check weight must be calibrated periodically
by an accredited calibration laboratory to an uncertainty no greater than one-third of the uncertainty required
of the balance. Also, the mass value of each check
weight must change by less than this amount between
successive calibrations. A control chart is used to
show any trends in the mass value and to determine
the re-calibration interval (typically three to five years).
Check weights must be treated with care if they are
to remain stable in mass value. Don’t let them get
dirty. Avoid exposing them to extremes of temperature
(moisture will condense on cold weights brought into a
plug). Keep check weights away from strong magnets.
Lift and place weights, don’t slide them. Be wary of
weights of unknown specification and home-made
weights – they may be too magnetic [6].
References and Bibliography
[1] C M Sutton, J E Robinson, M T Clarkson and G F
Reid, 2012, Balances and Weighing Workshop
Notes, (Lower Hutt: MSL, IRL).
[2] C M Sutton, J E Robinson and G F Reid, 2012,
Calibrating Balances, MSL Technical Guide 25,
http://msl.irl.cri.nz/training-and-resources.
[3] JCGM 100:2008 Evaluation of measurement data
- Guide to the Expression of Uncertainty in Measurement, available on the BIPM website at
http://www.bipm.org/en/publications under Guides
in Metrology.
Measurement Standards Laboratory of New Zealand  Fax 64 (0)4 931 3002  email: msl@irl.cri.nz  http://msl.irl.cri.nz/
Page 3
[4] Appendix G of Morris E C and Fen K M T, 2002,
The Calibration of Weights and Balances, Monond
graph 4: NML Technology Transfer Series, 2 ed.,
FINV and TINV.
[5] OIML R 111-1: 2004, Weights of classes E1, E2,
F1, F2, M1, M1-2, M2, M2-3 and M3, Part 1: Metrological and technical requirements, Organisation Internationale de Métrologie Légale.
[6] C M Sutton, 2004, Magnetic Effects in Weighing,
MSL Technical Guide 6,
http://msl.irl.cri.nz/training-and-resources.
Further Information
If you want to know more about balances and
weighing, contact MSL and book in for a Balances and
Weighing Training Workshop. See the MSL website
http://msl.irl.cri.nz/.
Prepared by C M Sutton and J E Robinson
Version 3, September 2012
MSL is New Zealand’s national metrology institute, operating within Industrial Research Limited under the authority of the
New Zealand Measurement Standards Act 1992.
Measurement Standards Laboratory of New Zealand  Fax 64 (0)4 931 3002  email: msl@irl.cri.nz  http://msl.irl.cri.nz/
Page 4
```