Sunteți pe pagina 1din 7

Accred Qual Assur (1999) 4:504-510

© Springer-Verlag 1999

Thomas Anglov Uncertainty of nitrogen determination


Inge M. Petersen
Jesper Kristiansen by the Kjeldahl method

Abstract The uncertainty of the largest contribution to the uncer-


Kjeldahl method for determination tainty came from volumetric equip-
of nitrogen in insulin was evalu- ment. Systematic uncertainty bud-
ated according the procedure de- gets such as the design presented
J. Kristiansen (lEI)
scribed in the Guide to the Expres- here facilitate the uncertainty eval-
The National Institute of Occupational sion of Uncertainty in Measure- uation process and makes it easier
Health, LerSI/l Parka lie 105, ment. The relative standard uncer- to compare uncertainty evaluations
DK-2100 Copenhagen, Denmark tainty of the method was found to performed by different analysts.
e-mail: jkr@ami.dk be 0.19%, compared to the relative
Tel.: +45-39-165200
Fax: +45-39-165201 intermediate precision experimen-
tally found to be 0.085%. The un- Key words Uncertainty budget·
T. Anglov, I.M. Petersen certainty components were organ- Uncertainty evaluation .
Department of Metrology, Novo Nordisk
A/S, Krogshl/ljvej 51, DK-2XXO Bagsvrerd, ized in Tables, which allowed an Uncertainty component·
Denmark easy overview and evaluation. The Traceability . Reference standards

Introduction Methods
Determination of nitrogen content plays a key role in Kjeldahl method
assigning values to insulin reference materials. The ref-
erence materials in question serve as reference stand- The nitrogen content of dry insulin was determined by
ards when measuring insulin in drug products. Thus, the Kjeldahl method, which consists of three steps, di-
the Kjeldahl nitrogen determination is a crucial link in gestion, distillation and titration. In brief, approximately
the traceability chain. The uncertainty budget for the 50 mg of the sample was weighed accurately and trans-
Kjeldahl method published in this paper has three ob- ferred to a digestion test tube. Concentrated sulphuric
jectives: Firstly, to estimate the uncertainty of results acid was added and the mixture was heated to 390°C
obtained by the Kjeldahl method; secondly, to identify for 4 h (digestion). The tube with the digested sample
steps in the analytical procedure that may be targets for was placed in the Kjeldahl apparatus, sodium hydrox-
improvement; and thirdly, to contribute a generally ap- ide and steam was added, and ammonia was distilled
plicable procedure for evaluating an uncertainty budget off. The ammonia-rich steam was condensed in the re-
for a chemical analytical method to the current litera- ceiving flask (distillation). The content of ammonia in
ture. The need for such schemes or procedures is ur- the receiving flask was determined by end-point titra-
gent as accredited laboratories in the near future will tion to pH 4.5 with 0.1 molll hydrochloric acid (titra-
be required to state their uncertainty of measurement tion). The normality of the acid (N HCh in mol/I) was de-
[1]. The uncertainty budget published in this paper is termined by titration of tris-(hydroxymethyl)-amino-
based on existing guidelines [2, 3]. methane (Tris). Blanks consisted of empty digestion
274 T. Anglov . I.M. Petersen' 1. Kristiansen

tubes treated as samples. Samples were measured in The factors 1Iy2 and 1/y3 account for the number of
duplicate and blanks in triplicate. replicate determinations (2 and 3, respectively). The
Let a (mg) denote the amount of sample. If b (ml) expression may be rearranged in order to emphasize
and c (ml) denote the volume of hydrochloric acid used the sensitivity coefficient of each uncertainty compo-
for titration of the blanks and the sample, respectively, nent:
then the relative nitrogen content of the sample (Ntota, )
can be calculated as: (N .)2 = ( Ntotal )2 ( )2 + ( N total )2 (b)2
U totdl y2 . (c _ b) u C y3 . (c _ b) u
N _ 14.01 g/mol· (c - b)· N HC1
total - (1)
a + (N to tal)2 u(a)2 + (N to tal)2 u(N HCI )2 (4b)
y2·a N HCI
The method was validated using two insulin drugs. The
relative standard deviation under intermediate preci- The standard uncertainties u(N HCI ), u(a), u(b) and u(c)
sion conditions was found to be 0.085%. are themselves composed from various uncertainty
components. These uncertainties were likewise ob-
tained from uncertainty budgets (Appendix Bl-4).
Uncertainty budgets

The uncertainty of Ntotal was estimated by combining


standard uncertainties of N HC ]' a, band c. Uncertainties Relative uncertainty variance contributions
were combined by using the rule of "error propaga-
In this paper the relative uncertainty variance contribu-
tion" [3, 4]. In general, when a result of measurement
(y) is determined from other quantities, the relation-
tions are used to illustrate the relative significance of
ship between y and the values of these quantities (input different uncertainty components [5]. The relative con-
estimates) can be expressed by a function, / [2]: tribution (rJ of an uncertainty component Xj to the
combined uncertainty, is defined here as the standard
y = f(Xb X2 ••• ,Xj ••• ,XN) (2) uncertainty variance of the component multiplied with
its squared sensitivity coefficient divided by the com-
where XI .•. Xj ...XN represents N input estimates. The un-
certainty of y (u(y)) is related to the uncertainty of the bined standard uncertainty variance [5]:
input estimates by the equation:

(3) (5)

where U(XJ2 is the standard uncertainty variance and


where xj=N HC ], a, band c. It follows from Eqs. (3) and
u(xJ the standard uncertainty of the uncertainty com-
(5) that L rj = 1 (assuming the absence of correlated in-
ponent number i, and where a/ is the partial deriva- put estimates).
aXj
tive of the function, f, with respect to the uncertainty
component number Xj. The partial derivative is often
called the sensitivity coefficient because it describes Results
how the measurement result varies with changes in the
value of the input estimates [2]. It should be noted that Uncertainty of Ntotal
Eq. (3) is an approximation that is valid if there are no
correlated input estimates [2]. The result of measure- The values of N HC ]' a, b, c and Ntotal as well as the re-
ment of the Kjeldahl method is N total as given by spective standard uncertainties u(N HCI ), u(a), u(b), u(c)
Eq. (1), and the expression for N total is the function/in and u(Ntotal) and the corresponding sensitivity coeffi-
Eq. (2). Thus, the following expression for the uncer- cients are given in Table 1. The combined standard un-
tainty of Ntotal is obtained: certainty u(Ntota') was estimated to 0.00023, corre-
sponding to a relative standard uncertainty of 0.19%.

7f
U sing an coverage factor k = 2 the result of the meas-
( U(C))2 (U(b))2
( U(Ntotal))2 = if + urement should be reported as:
N total (c-b)
(4a) Ntotal ± V(N total ) =0.1233 (±0.00046)
u(a))
+ (U(N HCI ))2 + ( £ where V(Ntotal) is the expanded standard uncertainty.
Details on the uncertainty budgets for u(N HCI ), u(a),
N HC1 a
u(b) and u(c) are given in Appendix B.
Uncertainty of nitrogen determination by the Kjeldahl method 275

Table 1 Values of uncertainty components, their standard uncertainty and sensitivity coefficient

Component Symbol Value Standard Sensitivity coefficient Reference


uncertainty

Normality of hydrochloric acid N Hcl 0.1 molll 1.1 x 10-4 molll (N ,o,al) = 1.23 llmol Appendix B1
N Hcl

Amount of sample a 50mg 0.059 mg (NTo,al) = 1.74 X 10-3 mg- I Appendix B2


V2'a
Volume of titrant used on blank sample b 0.1 ml 4.0 x 10-3 ml ( N'o'a' )=1.fl2XI0- 2 ml- 1 Appendix B3
tHe-b)
Volume of titrant used on the sample e 4.5 ml 7.2x10 3 ml ( N'o'a' )=1.9HXI0- 2 ml- I Appendix B4
V2·(e-b)
Result of measurement Niota' 0.12329 0'()()()23 Eq. (4b)

Relative contributions to the uncertainty cludes the contributions from uncertainty of the vol-
ume of the titrant and from uncertainty of the tempera-
The relative contributions (r;) from U(N HCI ), u(a), u(b) ture of the titrant (Appendix Bl, 3 and 4). Weighing
and u(e) to the combined standard uncertainty variance (Fig. 2) includes first and second weighing of both Tris
u(Ntotal)2, are shown in Fig. 1. The two largest contribu- base and the sample (Appendix Bl and 4). Digestion
tions come from U(NHCI) and u(e). Both contribute (Fig. 2) includes all uncertainty components denoted in
35-38% to the combined standard uncertainty var- digestion in Appendix B3 and 4. Lastly, Tris-purity
iance. The uncertainty budget for U(NHCI) (Appendix (Fig. 2) denotes the uncertainty component associated
Bl) shows that the uncertainty of this component is with the purity of the Tris base evaluated in Appendix
composed mainly of uncertainty contributions from the B1. From Fig. 2 it can be seen, that the largest uncer-
temperature of the titrant, the weighing of the Tris base tainty contribution comes from the use of volumetric
and the volume of the titrant. The primary contribu- equipment, i.e. burettes used for titration of hydro-
tions to a(e) (see Appendix B4) come from the uncer- chloric acid and for titration of samples and blanks.
tainty of digesting the sample (which consists of the un- This uncertainty component is composed of the preci-
certainty of the amount of sample transferred to the di- sion of titration and of the uncertainty of the tempera-
gestion tube) and from the uncertainty of the volume of ture. The latter factor influences the volume of titrant
the titrant. used for titration. The second largest contribution
In Fig. 2 the relative contributions to the uncertainty comes from weighing (of Tris base and of the sam-
variance are summarized for various types of analytical ple).
steps and equipment used in the Kjeldahl method. The
contribution from volumetric equipment (Fig. 2) in-

~
c
0
~ i
.c
c
0
i.c i
..
;:
c
0
8.:
!J
u
..:
::

Fig.2 Relative contributions from volumetric equipment, weigh-


Fig. 1 Relative contributions from N He," a, band e to the com- ing, digestion and the purity of the Tris base to the combined
bined standard uncertainty variance u(N'o,al)2 standard uncertainty variance
276 T. Anglov . I.M. Petersen' J. Kristiansen

It was not found necessary to consider the correla-


Discussion tion between input estimates because the contribution
The uncertainty budget presented in Table 1 indicates from covariances was judged to be minor in the uncer-
that a typical result obtained by the Kjeldahl method tainty budget presented above. Examples are c and b in
(as described in this paper) will have a relative standard Eq. (1) which are determined using the same burette.
uncertainty of around 0.19%, which is approximately This means that uncertainties of c and b originating
twice as large as the experimentally derived relative from, e.g. long-term variations of volume (delivered by
standard deviation estimated under intermediate preci- the burette), wear of the burette, etc. are positively cor-
sion conditions. This is not surprising as some of the related. In other words, the uncertainty of the differ-
uncertainty components included in the uncertainty ence (c - b) tends to be unaffected by these effects. The
budget are difficult to evaluate experimentally, e.g. pur- same reasoning can be used for m1 and m2, which are
ity of the Tris base, accuracy of the pH meter, etc. determined for calculation of a (Appendix B2).
The uncertainty evaluated for the Kjeldahl method The structure of this uncertainty budget for the Kjel-
should be propagated to the uncertainty of the insulin dahl method may be used as a paradigm for other ana-
reference standards as well as to the analytical methods lytical methods. However, it must be understood that
that are calibrated with these standards. Failure to con- an uncertainty budget is designed for a specific method,
sider the uncertainty of the reference standards or of and is therefore only valid for a set of specified analyti-
the analytical method, or to base the uncertainty on less cal conditions (including the analytical method, type of
encompassing estimates (e.g. intermediate precision), sample and its concentration level). For highly auto-
may lead to an excess number of nonconforming results mated (i.e. computerized) analytical methods, the result
when evaluating the fulfilment of acceptance criteria of measurement is often calculated by specialized soft-
for insulin drug products. ware, and this makes the interdependency of various
The present uncertainty budget demonstrates that analytical steps less obvious. For such systems other ap-
the Kjeldahl method may be improved. The most proaches for making uncertainty budgets must be ap-
promising target for improvement is the titration. Ac- plied [5, 6].
cording to the analysis of the uncertainty contributions,
this can be accomplished by improving the temperature
control during titration or by improving the accuracy of Conclusions
the volume of titrant. The latter goal may be achieved
by using more accurate titration equipment, or, at least A structured evaluation of uncertainty components was
partly, by increasing the number of replicate determi- applied for the Kjeldahl method for nitrogen determi-
nations. Increasing the number of replicates will im- nation. The design offers a systematic approach for
prove the precision, but will not influence the trueness making uncertainty budgets for analytical chemical
of the volumetric equipment, nor does it compensate methods, and it may facilitate comparison of uncertain-
for long-term fluctuations that may affect the equip- ty budgets made by different analysts or laboratories.
ment. The evaluation of the contribution from individual or
The uncertainty budget in this study was organized grouped uncertainty components made it possible to
to make it easy to review the evaluation of individual suggest specific improvements of the analytical meth-
uncertainty components. The uncertainty budget is or- od.
ganized in three evaluation steps: In the first step, the
uncertainty is evaluated based on the input estimates
that are used directly in the calculation of Nlolal (i.e. Appendix A: Definitions, abbreviations and equations
N HCb a, band c in Eqs. 1 and 4b) (results in Table 1). In
the second step, the uncertainty of each of the input 1. Definitions and abbreviations
estimates (i.e. N HCb a, b and c) is evaluated and calcu-
lated. Correction factors are applied to correct for ef- Terms and definitions used in accordance with GUM
fects such as the influence of temperature on the vol- and VIM [2, 7].
ume, inaccuracy of the pH meter, etc. (equations and
results in Appendix B). Where necessary, a third step a: mg sample used in the measurement
of uncertainty evaluation was applied. An example is b: ml titrant (hydrochloric acid) used on the blank
air bubbles in the burette, water in the Tris base and sample (average of three determinations)
other effects that are evaluated and included in the un- c: ml titrant (hydrochloric acid) used on the sam-
certainty of qconlT (Appendix B1). Generally evalua- ple
tions in the third step should only be carried out if the k: Correction factor 14.01 g/mol containing the
uncertainty components contribute significantly to the atomic weight of nitrogen
combined uncertainty. MPE: Maximum permissible error
Uncertainty of nitrogen determination by the Kjeldahl method 277

Table 81 Estimation of the uncertainty of NHC1


No. Input Symbol Value Stated or Type Standard Sensitivity coefricient Contribution to Reference
quantiliy (x,) of input evaluated (AlB) and uncertainty ar the standard
estimate uncer- type of u(x,) ax, uncertainty
tainty distribution of the output
estimate
ar
u(x,)~
ax,

various
contri-
q~ulltr
lxlO-4, B
rectangular
CX~)-4)=5.xX 10-' N lln =0.1 molll
qC~lnlr
5.X x 10-6 molll Own estimate

butions'
2 MPE of qrll '" Oh Instrument
the pH- specification.
meter own estimate
3 Purity of qpu, 0.9989 (UKW)SC B 2.9 x 10-' molll Supplier
Tris (99.X9%) liJ- .x 0
0.005 _ 2 9 1 -4 Nlln=O.1 molll
qpur
certificate
4 First m, 11K) mg 0.1 mg B O.lmg 5.9 x 10-' molll Instrument
rectangular Nlln = 1.02 x 10-' moll(mg·l) specification
weighing
of Tris base 7~
5.8x 10-" mg
m,-III,

5 Second m, 2mg 0.02mg B 0.02 mg 1.2 x 10-' molll Instrument


weighing rectangular yr= ~= 1.02 x 10-' moll(mg·l) specification
ln2-n1 1
of Tris base 1.2 x 10-' mg
6 Volume of v 8.06 ml 0.004ml" A 0.004ml Nlln 5.0xI0-' molll Instrument
titrant normal - - = 0.012 moll(L'ml) qualification
v
report
Temperature O.lX1l2 6.9 x 10-' molll Own estimate
7 (1.2~~0-')=6.l)XIO-4 Nlln =0.1 molll
C
qlcmp
B
of titrant rectangular r J q'cmr Laboratory
temp.
15-25°C
Molar mass M ...,;, 12l.l4 ",Og/mol' - Own estimate
of Tris base glmol
u (XN",) =Y(5.8 x 10 ")' + (2.9 x 10 Sf + (5.9 x 10 'f + (1.2 x 10 ')'+ (5.0 x 10 ')'+(6.9 x 10 ')' = l.l x 10 -4 molll

a Contributions from: The water content of the silica gel; lack of C The stipulated purity of (hydroxymethyl)-aminomethane (Tris)

complete drying of the Tris base; variation in the time of drying; is 99.H9% with an uncertainty of (l.05%
variation in the time of dissolution of the Tris base; the hygrosco- d A standard deviation of 0.004 mL was found experimentally
py of the Tris base; air bubbles in the burette: Combined uncer- C Water density at 15°C is 0.99913 glmL, at 20°C 0.99H23 glml and

tainty estimated to 0.01 mg Tris base or 0.01 % of approximately at 25°C 0.99707 glmL. It is assumed that the temperature in the
100 mg Tris base laboratory (and the temperature of the acid) in average is 20°C,
h The pH-meter specifications stipulate a MPE of 0.02 pH units. but in worst case may vary ±5°C. The largest change in water
The form of the titration curve indicate that an uncertainty of 0.02 density from 20 to 25°C is 0.00116 g/ml, or 0.12%
pH corresponds to a negligible uncertainty in the volume of ti- f The uncertainty of the molar mass is assumed to give a negligi-
trant. Thus, the value of this uncertainty component is not in- ble contribution to the combined uncertainty
cluded in the uncertainty budget

Table 82 Estimation of the uncertainty of a


No. Input quantity Symbol Value of Stated or Type Standard Sensitivity Contribution Reference
(x,) input evaluated (AlB) uncertainty coefficient to the standard
estimate uncertainty and type of u(x,) ~r uncertainty
distribution ax, of the output
estimate
ar
u(x,)~
ax,

First weighing III, 52mg 0.1 mg B 0.1 mg 5.8x 10-' mg Instrument


of sample rectangular y1 specifications
2 Second weighing 2mg 0.02mg B 0.02 mg 1.2 x 10-' mg Ins(rument
vr
III,
of sample rectangular specifications

u(a)=Y(5.8XIO ')'+(1.2xlO ')'=0.059mg


278 T. Anglov . I.M. Petersen' J. Kristiansen

Table 83 Estimation of the uncertainty of b


No. Input Symhol Value of Stated or Type Standard Sensitivity Contrihution Reference
quantity (x,) input evaluated (AlB) uncertainty coefficient to the standard
estimate uncertainty and type of u(x,)
i!L uncertainty
distrihution ax, of the output
estimate
It (x,) af
ax,
Digestion' qJip.cst N/A N/A N/A N/A N/A Digestion not
performed
2 Distillation" qdistil Own estimation
3 Volumen of v 0.1 ml O.(X)4 ml A O.!X)4 ml e 4 x 10-' ml Instrument
titrant normal !!. = 1 qualification
v
report
4 Temperature qtcmp 1.2x 10-'" B
_b_=O.1 ml 6.9xlO-' ml Own estimation
of titrant rectangular qtcmp
Lahoratory
temp. 15-25°C
5 MPE of the qpH Instrument
pH meter specifications.
Own judgement

a No Kjeldahl mixture or acid was used. The digestion was not laboratory (and the temperature of the acid) on average is 20°C,
performed but in worst case may vary ± 5°C. The largest change in water
h Addition of water and NaOH; distillation time; leaks; incom- density from 20 to 25°C is 0.001 Hi glml, or 0.12%
plete transmission of NH3 to the receiving vessel; temperature of C The pH-meter specifications stipulate a MPE of 0.02 pH units.

the vapour: All components are assumed to be negligible The form of the titration curve indicates that an uncertainty of
C A standard deviation of 0.004 ml was found experimentally 0.02 pH corresponds to a negligible uncertainty in the volume of
d Water density at 15°C is 0.99913 glml, at 20°C 0.99823 glml and titrant. Thus, the value of this uncertainty component is not in-
at 25°C 0.99707 g/m\. It is assumed that the temperature in the cluded in the uncertainty budget

Table 84 Estimation of the uncertainty of c


No. Input Symhol Value of Stated or Type Standard Sensitivity Contrihution Reference
quantity (x,) input evaluated (AlB) uncertainty coefficient to the standard
estimate uncertainty and type of u(x,) af uncertainty
distrihution ax, of the output
estimate
u(x,)~
at"
ax,
Digestion' (Um B c 5.2 X 10-3 ml Own estimation
qdigcst
rectangular (O~2)= 1.2 x 10-' --=4.5ml
qdigcst

2 Distillation b qJistii Own estimation


3 Volume of v 4.5 ml O.(Xl4 ml A 0.004 ml' 4.0x 10-' ml Instrument
titrant normal ~= 1 qualification
v
report
4 Temperature (W1l2 B c 3.0x10-3 ml Own estimation
of titrant
qtemp
rectangular (().~12) =6.9 x 10-' --=4.5 ml Lahoratory
qJistil
temp. 15-25°C
5 MPE of the qpll zO ml" Instrument
pH meter speci[ications.
Own judgment
u(c)=V(5.2xlO ')'+(4xlO .l)'+(3.0xlO .l)2=7.22x 10-' ml

"The uncertainty components: Amount of catalysator, the C A standard deviation of O'()04 ml was found experimentally
amount of Kjeldahl mixture and H 2S0 4 , block temperature, di- d The pH-meter specifications stipulate a MPE of 0.02 pH units.
gestion time and boiling are assumed not to contribute signifi- The form of the titration curve indicates that an uncertainty of
cantly to the uncertainty. Transfer of sample to digestion vessel is 0.02 pH corresponds to a negligible uncertainty in the volume of
assumed to contribute with 0.1 mg or 0.2% titrant. Thus, the value of this uncertainty component is not in-
h Addition of water and NaOH; destillation time; leaks; incom- cluded in the uncertainty budget
plete transmission of NH3 to the receiving vessel; temperature of
the vapour: All components are assumed to be negligible
Uncertainty of nitrogen detennination by the Kjeldahl method 279

N: Number of quantities or uncertainty compo- rection factors for the water content of Tris, the accura-
nents cy of the pH meter, the temperature of the titrant, and
N HC (: The normality of the hydrochloric acid the purity of Tris (Table BI):
N tota (: Content of nitrogen in the sample (average of 2
samples)
qi: Correction factors
u(x;): Standard uncertainty of the input estimate Xi
U(y): Expanded uncertainty of y
u(y): The combined standard uncertainty of y 2. Estimation of the uncertainty of a
Xi: Input estimates
y: Output estimate (the result of a measurement) The input estimate a is obtained from weighing the
aJ. Sensitivity coefficient for Xi
sample (m2) and sample cup alone (m(), i.e.
aXi a = J(x;) = m2 - mi' Thus, the standard uncertainty of a
is given by the equation u,7 = u;', + u,;", (Table B2).

2. Equation for the standard uncertainty


3. Estimation of the uncertainty of b
In addition to Eq. (3) the following equations evaluat-
ing the standard uncertainty were applied: If the input In evaluating the uncertainty of b it was assumed that
estimates Xi in Eq. (2) is related to y only by multiplica- the volume of titrant used for titration of a blank de-
tion's and divisions, Eq. (3) can be simplified to [2, 4]: pends on factors associated with digestion, distillation,
pH measurement and the assessment of the volume of
(AI) titrant used. The combination of uncertainty compo-
nents was based on the following relation between the
with sensitivity coefficients given by: volume (b) and the various factors:

aJ y (A2)
aXi Xi where v is the estimated volume of titrant read from
If the input estimates Xi is related to y by additions and the burette and all the correction factors (qi) are equal
subtractions, Eq. (3) becomes: to I (Table B3).

U(y)2 = U(X()2 + U(X2)2 + ... + U(XN)2 (A3)


with all sensitivity coefficients equal to 1. 4. Estimation of the uncertainty of c

The uncertainty components and equations used are


Appendix B the same as for b (Appendix B3), i.e.

1. Estimation of the uncertainty of N He(


where v is the estimated volume of titrant read from
N Hc( is calculated from the amount of Tris base used, the burette and the correction factors (q;) are all equal
the molar mass of Tris, the volume of titrant, and cor- to I (Table B4).

References

1. International Standard ISO/IEC 17025 3. EURACHEM (1995) Quantifying un- 6. Hansen AM, Kristiansen J, Nielsen JL,
(199X) General requirements for the certainty in analytical measurement, Byrialsen K. Christensen JM (1990)
competence of testing and calibration 1st edn. EURACHEM Talanta 50: 367-379
laboratories (Draft). International Or- 4. Miller JC, Miller IN (1993) Statistics 7. ISO (1993) International vocabulary of
ganization for Standardization. Geneva for analytical chemistry. 3rd edn. Ellis basic and general terms in metrology
2. BIPM, IEC. IFCC, ISO, IUPAC. Harwood, New York (VIM), 2nd edn. International Organi-
IUPAP. OIML (1993) Guide to the 5. Kristiansen J. Christensen JM, zation for Standardization, Geneva
expression of uncertainty in measure- Nielsen JL (1996) Mikrochim Acta
ment. International Organization for 123:241-249
Standardization. Geneva

S-ar putea să vă placă și