Comparative Measurements on an Electronic [603464]

Comparative Measurements on an Electronic
Weighing Instrument
Adriana V âlcu, Sterică Baicu
Mass Laboratory
National Institute of Metrology (INM)
Bucharest, România
e-mail: [anonimizat]
sterica .baicu @inm.ro Alina Taină, Roberta Todor
ROLAB Association
Bucharest, România
email; [anonimizat]
roberta_t [anonimizat]

Abstract — Many times was heard the dilemma: “To measure or
not to measure, that is the question". But, when you measure, a
question arises: how accurate/reliable are your measurements
results so that the client should have the confidence that you
measure in the right way? No matter how carefully you perform
internal checks, systematic errors often can go unnoticed. For
this reason, the simplest way to check the quality of your
measurements is to participate in proficiency testing scheme
(PTS) / interlaboratory comparison (ILC). Moreover, ISO/IEC
17025 requires “that laboratories have quality control procedures
for monitoring the validity of tests and calibrations undertaken”.
It is therefore important that you should regularly compare you r
measurement results with those of other laboratories so, to attend
in intercomparisons. In this way, this paper describes and
analyzes the comparative results and uncertainties associated
obtained by a reference laboratory and other six participating
laboratories, appearing with codes that have been assigned
during the comparison. The proficiency test item consisted in a
NAWI (non -automatic weighing instrument ) having Max 6 kg
and resolution of 10 mg. The measurements were performed at
Mass laboratory of INM which was designa ted as Pilot (LP) and
Reference Laboratory (LR). The determinations began in March
2015 and concluded in August 2015. The goal of interlaboratory
comparison was to provide verification of each participating
laboratory’s measurement capability by obtaining a
measurement that agrees with the LP/LR. The results are
analyzed using En (normalized errors) values.
Keywords: electronic balance; error of indication; normalized
error; inter -laboratory comparis on/proficiency testing .
I. INTRODUCTION
The aim of an interlaboratory comparison is to verify the
competence of accredited or non -accredited laboratories,
including verification of the reported measurement
uncertainties, whenever possible. Participation in a proficiency
testing can validate the participating laboratory’s measurement
method, technical training, t raceability of standards, and
uncertainty budgets. This paper reports the results obtained by
six metrology laboratories in a proficiency testing scheme .
The subject of the comparison (the proficiency test item)
was an electronic w eighing instrument with Max 6 kg and
resolution d = 10 mg. The aim of the comparison was not only to compare
measurement results of the participant with those of PL/RL, but
also to analyze measurement uncertainty, the choice of
calibration points and the content of calibration ce rtificates.
The result is considered successful, if the value of the
normalized error is -1 ≤ En ≤ 1. In this case the participating
laboratory agrees with the reference value within the stated
uncertainty. If a laboratory has any unsuccessful results, i.e En
< -1 or En > 1, it is expected that the laboratory investigates the
reason for the disagreement and implements corrective action.
II. MEASUREMENT INSTRUCTI ONS
The following measurement instructions were provided to
participants:
– calibration will be performed over the full weighing range
from zero to the maximum capacity Max;
-the calibration points are established by each participating
laboratory;
– for acclimatization, it was proposed that each laboratory
bring in advance the weights at LP / LR ;
– it wa s not imposed a particular method for application of
loads (ascending, descending or both);
– after completion of the measurements, in a maximum
period of 10 working days, each participating laboratory shall
submit the results to the scheme coordinator, in the required
format of the proficiency test protocol;
– metrological and technical characteristics of the weighing
instrument were provided before starting the proficiency
testing.
III. PARTICIPATING LABORATORIES
For the presentation of measurement results w ere used
codes (B to G) attributed to participating laboratories by PL /
LR, which had the code "A".
Initially, 6 laboratories announced their participation in the
comparison. Subsequently, one of these (laboratory marked
with "F"), was excluded from the a nalysis of PT because the

submitted documentation contained discrepant information
between Annexes, having also many miscalculations. For this
reason, the results could not be used in analyzing the
normalized errors within the prescribed limits. In the dra ft
report, the laboratory F appears as a participant, but not in the
analysis of the results.
Laboratory coded with "E” performed the calibration of the
balance, but has not sent the final results and announced its
withdrawal from the PT.
IV. TYPE OF COMPARISO N SCHEME , PLANNING OF
PARTICIPATING LABORATORIES
A. Measurement scheme of PT

Figure 1. Comparison s cheme
B. Calibration data for participating laboratories

In Table I is presented the data when the laboratories
calibrated the balance.
TABLE I. THE DATA WHEN EACH PARTI CIPANT LABORATORY
PERFORMED THE CALIBR ATION

V. MEASUREMENTS BY INM
The mass laboratory of INM is the national standard
laboratory for mass in Romania . The measurements results are
traceable to International System of Units (SI).The mass unit
comes from BIPM. For the calibration of the balance, PL/RL used direct comparison method with standard weights having
nomina l mass and accuracy class presented in the Table II.
TABLE II. THE STANDARD WEIGHTS USED BY PL /RL IN THE
CALIBRATION

Set of weights OIML
Class Uncertainty Calibration
certificate
500g and 500 mg,
No. 51 E2 (0,15…0,005) mg 02.01 -1351/201 5
(1…5) kg, No.
G045982 E2 (0,3…1,6) mg 02.01 -118/2015

The test loads used for determining the errors of indication
were: 0, 0.5g, 500g, 1000g, 1500g, 2000g, 2500g, 3000g,
3500g, 4000g, 4500g, 5000g, 5500g and 6000g. Repeatability
was determined at 3000 g and 6000 g and eccentr ic errors at
2000 g.
Each load was placed only once, increasing by steps with
unloading between the separate steps .
The errors of indication and uncertainty associated to these
errors were determined accord ing to formulae (1) and (2), [1],
namely:
Ej = Ij – mrefj 

with: Ij the indication of the balance
mrefj is conventional mass of the reference weights

 
 
u(I)2
The significance of he terms fr om formula (2) is presented
in the Figure 2 [2].

Figure 2. Ishikawa diagram of uncertainty components in the calibration of
the balance Laborator y’s code Calibration Data
B 19.03.2015
C 09.06.2015
D 09.07.2015
E 21.07.2015
F 04.08.2015
G 11.08.2015
,2 2 2 2 max
0 ( ) 2 ( ) /12 /12 ( ) ( )
2 3ecc i
L mref
eccI
U E u E d d s I I u m
L
     

VI. RESULTS OF THE COMPARISON
A. Method and standard weights used for calibration
For loading the balance, laboratories used the follow ing
methods:
– laboratories B, G and E have opted to load the balance
only once, increasing, by steps with unloading between the
separate steps, corresponding to the majority of uses of the
instruments for weighing single loads [ 1].
– laboratories C, D, a nd F have performed both, increasing
and decreasing loading of the balance. Laboratory F performed
also the tare test, although it was not necessary for the
calibration of NAWI.
Table III indicates the accuracy classes of standard weights
used by each part icipating laboratory for calibration of NAWI.
TABLE III. THE STANDARD WEIGHTS USED BY PARTICIPANTS
Laboratory Nominal mass OIML Class
B (5kg…0,5g) E2
C (5kg… 0,5g) E2
D (5kg… 0,5g) F2
E 500 and 1000g E2
(5…2) kg F2
F 0,5 g E2
(5…1) kg F1
G (5kg… 0,5g) F2
B. Measurement uncertainty
Measurement uncertainty was estimated according to [1],
with the following observations:
– uncertainty ass ociated to limited resolution was estimated
only once by the laborator y D and laboratory F did not
introduce this component in the calculation of standard
uncertainty ;
– laboratories D and F have not introduced in uncertainty
budget a component associated to air buoyancy ;
– repetability was miscalculated by laboratory F and from
uncertainty budget s of laboratory D and F it is not clear if
repeatability was included in the calculation;
– laboratory D has not introduced in the calculation of
standard uncert ainty the component associated to eccentricity;
– laboratory F miscalculated the component associated to
load eccentrically placed; in addition, in the calibration
certificate are presented results for two loads, although from
the primary data, results th at only at 2 kg was performed this
test;
– laboratory C mistyped the numerical value of repeatability
in the table containing uncertainty budget ; also, laboratory used
for all test loads the same value for uncertainty associated to
effect of eccentricity; the value for eccentricity error it should
be given in modul.
– although laboratories encoded with C, D and F performed
accuracy test with increasing and decreasing loads, in the uncertainty budget it is not found a component associated to
hysteresis.
In T able IV are shown the components of uncertainty
reported by each participating laboratory.
TABLE IV. UNCERTAINTY COMPONENT S REPORTED BY EACH
PARTICIPATING LABORA TORY

C. Results in tabelar form
A tool often used in analyzing the results from
interlaboratory comparis ons is the normalized error En, which
takes into account both the result and its uncertainty. The
normalized error En is given as :

2 2
ref labref lab
nU Ux xE
 
The subscript “lab” refers to the participating laboratory
and “ref “refers to INM (reference laboratory).
The participants’ results were analyzed in two directions ,
because between the first laboratory determinations and the
next participant, it was f ound an alteration of the balance
characteristics. It can be considered that one of the causes of
this change could be the long time between the two calibrations
(see Table I). Thus, the laboratory B performed the calibration
on 03/19/ 2015, at a distance o f about three months from the
rest of the group of participants. For this reason, only for
laboratory B, it was chosen to calculate the normalized error
using the data of reference laboratory obtained in the same day.
For the rest of participants, the refe rence values were calculated
from the average of all LR results (without data from
03/19/2015). It was found that using average values of LR that
contained data from 03/19/2015, the final results would have
been distorted for both the laboratory B and LR.
Table V presents the errors of indication, " E" and the
expanded uncertainty " U" (k = 2) associated with these errors
reported by laboratory B and laboratory A (reference)
performed on 03/19/2015. Table VI presents the errors of
indication, " E" and expanded uncertainty " U" (k = 2)
associated with these errors reported by the rest of participating
laboratories and reference laboratory A. The results are
presented exactly in the form that was sent by the participants .

TABLE V. THE ERRORS OF INDICAT ION “E” AND EXPA NDED UNCERTAINTY “U” REPORTED BY LABORATO RIES B AND A

Load
g Laborator y
A B
E
mg U
mg E
mg U
mg
0 0 12
0,5 0 12 0 30
500 -20 12
1000 -10 12
1500 -30 13 -20 30
2000 -19 13
2500 -9 14
3000 -29 15 -30 30
3500 -49 30
4000 -71 31
4500 -91 32 -100 60
5000 -130 33
5500 -160 34
6000 -200 34 -200 60
TABLE VI. THE ERRORS OF INDICAT ION “E” AND EXPANDED UNCERTA INTY “U” REPORTED BY LABORATO RIES C, D, E, F, G AND A

Load
g Participating Laboratories
A C D E F G
E
mg U
mg E
mg U
mg E
mg U
mg E
mg U
mg E
mg U
mg E
mg U
mg
0 0 16,5 0 8 – – – – 0 20
0,5 -4,0 16,5 5 6 -10,12 8 – – – – -10 20
10 -4,2 16,6 5 14 – – – – – – – –
50 -5,4 16,7 – – -10,5 9 – – – – – –
100 -6,8 16,9 -10 6 -20,2 9 – – – – – –
500 -18,0 18,7 -10 6 -20,5 14 – – – – – –
1000 -8,0 18,7 0 28 -8 22 – – – – -10 20
1500 -24,0 20,8 – – – – – – – – – –
2000 -13,4 23,3 -1 25 -3 45 – – – – -20 30
2500 4,6 26,2 – – – – – – – – – –
3000 -11,4 29,3 -1 25 -1 49 – – – – -20 40
3500 -9,4 34,2 – – – – – – – – – –
4000 -14,5 37,4 -11 28 -35 63 – – – – -20 60
4500 -30,5 40,7 – – – – – – – – – –
5000 -52,4 44,1 -65 25 -23 109 – – – – -20 70
5500 -74,4 47,5 – – – – – – – – – –
6000 -84,4 50,1 -80 25 -41 111 – – – – -50 80

D. Normalized error s of the reported values
In Tables VII a and b are presented the normalized errors
En of the reported values compared to reference values for all the loads where the balance was calibrated by each
laboratory .
TABLE VII. NORMALIZED ERRORS EN VALUES OBTAINED FOR EACH LABORATORY
Table VII a Table VII b

E. Results in graphic al form
Figure 3 presents the results of laboratory A (LR) for all
six determinations. From this graph it can be seen how far
are the results of laboratory A (LR) performed on
03.19.2015 , from the rest of determinations.
Figure 4 presents the results obt ained by laboratory A
(LR) for the five determinations (when was excluded the determination performed in 03.19.2015). These five
determinations were used for calculation of the normalized
errors for laboratories C, D and G.
Figures 5…8 presents the resul ts obtained by laboratories
C, D and G.
Figure 9 contains a centralization of En values.

Figure 3. Results obtained by laboratory A (reference) for all six
measurements

Figure 4. Results obtained by laboratory A (reference) without data from
19.03.20 15
Load
g Laboratories
C D G
0 0,0
0,5 0,5 -0,3 -0,2
10 0,4
50 -0,3
100 -0,2 -0,7
500 0,4 -0,1
1000 0,2 0,0 -0,1
1500
2000 0,4 0,2 -0,2
2500
3000 0,3 0,2 -0,2
3500
4000 0,1 -0,3 -0,1
4500
5000 -0,2 0,2 0,4
5500
6000 0,1 0,4 0,4 Load
g Laboratory
B
0
0,5 0,0
500
1000
1500 0,3
2000
2500
3000 0,0
3500
4000
4500 -0,1
5000
5500
6000 0,0
Laboratory A (Reference), data from all six measurements
-250-200-150-100-50050
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000Load [g]Error of indication[mg]
19.03.2015 09.06.2015
09.07.2015 21.07.2015
04.08.2015 11.08.2015
– U + U
Average of reference values (laboratory A)
Laboratory A (Reference) without data from 19.03.2015
-250-200-150-100-50050
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000Load [g]Error of indication [mg]
09.06.2015 09.07.2015
21.07.2015 04.08.2015
11.08.2015 Average of reference values (laboratory A)
– U + U

Figure 5. Results obtained by laboratory B Figure 6. Results obtained by laboratory C
Figure 6.

Figure 7. Results obtained by laboratory D Figure 8. Results obtained by laboratory G
Centralization of En values
0 0,5 10 50 100 500 1000 1500 2000 3000 4000 4500 5000 6000 g
Laboratory B Laboratory C
Laboratory D Laboratory G
-1 +1

Figure 9. Centralization of En values for each participant labora tory
Laboratory B
-300-250-200-150-100-50050
0
1000
2000
3000
4000
5000
6000
Load [g]Error of indication [mg]r
Laborator A
Laboratory B
– U
+ U
Average of reference values (laboratory A)
Laboratory C
-200-150-100-50050
0
1000
2000
3000
4000
5000
6000
Load [g]Error of indication [mg]]
Laborator A
Laboratory C
– U
+ U
Average of reference values (laboratory A)
Laboratory G
-200-150-100-50050100
0
1000
2000
3000
4000
5000
6000
Load [g]Error of indication [mg]
Laborator A
Laboratory G
– U
+ U
Average of reference values (laboratory A)
Laboratory D
-200-150-100-50050100
0
1000
2000
3000
4000
5000
6000
Load [g]Error of indication [mg]
Laborator A
Laboratory D
– U
+ U
Average of reference values (laboratory A)

VII. CONCLUSIONS
Six mass calibration laboratories announced their
participation in a comparison of a NAWI, having Max 6000g
and d = 10mg.
Subsequently, one of these (laboratory marked with "F"),
was excluded from the analysis of PT because the submitt ed
documentation contained discrepant information between
Annexes , having also many miscalculations. For this reason,
the results could not be used in analyzing the normalized
errors within the prescribed limits. In the draft report, the
laboratory F appea rs as a participant, but not in the analysis
of the results.
Laboratory coded with "E” performed the calibration of
the balance, but has not sent the final results and announced
its withdrawal from the PT.
The results for the rest of participating laboratories (B, C,
D and G) were in agreement with the reference values of
LR/LP.
It should be make a distinction between the final results
of the participants in the sense that in some cases they do not
correctly reflect the la boratory's capability. So:
– Laboratory C used the same value for uncertainty
associated to eccentric effect. This contributed to an
undervaluing of the combined uncertainty, given that the
error of eccentricity was 29 mg. – Laboratory D has not introduced all the components in
uncertainty budget;
– Laboratories D and G used weights of accuracy class F 2,
although for this weighing instrument it was appropriate to
use at least class F 1 weights. For this reason , standard
uncertainty of indication errors is higher.
It was made a suggestion for laboratory F to participate in
the following training courses regarding calibration of
NAWI .
In our country, until recently, only in the field of
calib ration of weights were requests for organization of inter –
laboratory comparisons.
Proficiency testing in the field of calibration of NAWI is
a newer subject, being an important requirement for both,
laboratories accredited for the calibration of non -autom atic
weighing instruments and accreditation bodies.
REFERENCES
[1] EURAMET, cg -18 “Guidelines on the Calibration of Non -automatic
Weighing Instruments ”, version 4.0 /201 5.
[2] Adriana V alcu, “ Between Parallel Mirrors – An Electronic Weighing
Instrument ”, Proceedings of t he 9th International symposium on
advanced topics in electrical engineering , Bucuresti, 2015.
[3] SR EN ISO/CEI 17043 , “Conformity assessment — General
requirements for prof iciency testing ”, 2010 .