a division of under the

ITDI DOST

NML
Determination of measurement
uncertainty in the calibration of
piston-operated pipette

[Link]
Metrology Building
DOST Compound
Gen. Santos Avenue
Bicutan, Taguig City
November 09, 2021
Measurement Uncertainty
• Real or actual measurements are not performed under perfect conditions.
Thus, the measured values differ from the true value of measurand.
true value

measured valuesmeasured values

“measured values are estimates of the true value”

• The average value of the measured values is the best estimate of the true value, 𝒚
• Thus, there is always some doubt on the value of the measurand in every
actual measurement.
• The size of the margin of this doubt is called uncertainty, 𝑼. Uncertainty is a
measure of the quality of the measured value.
2 [Link]
Measurement Uncertainty
• Uncertainty ≠ Error
• Systematic error (deviation) or measurement bias is the difference between the
measured and nominal quantities
Example: Measured value is 5.08 uL ; Nominal value is 5.00 uL
Error: 5.08 uL -5.00 uL = 0.08 uL
• Correction is the opposite of error
Example: from the example above, the correction value is -0.08 uL
-0.08 uL is the value needed to adjust the measured value back to the
nominal value.
5.08 uL + (-0.08 ul) = 5 uL

3 [Link]
Estimation of Measurement Uncertainty
Perform calibration and record raw data

Estimate Measurement Uncertainty

Develop model equation

Compute for water density & air density
Compute for the actual water
contained/dispensed using the model equation
Identify sources of uncertainty
Compute for the standard uncertainty Compute the expanded uncertainty
contributed by each source Determine the coverage factor &
Compute for the sensitivity coefficient of each confidence level
source Sum up all the uncertainty contributions
Compute the uncertainty contribution of each to obtain the combined standard
source uncertainty

4 [Link]
Model Equation
1 𝜌𝑎
Vo= 𝑚 × (𝜌 −𝜌𝑎)
× 1−
𝜌𝑏
× 1 − 𝛾 𝑡𝑑 − 25 ℃
𝑤

Vo = volume, at the reference temperature

𝑚 = mass of water obtained dispensed or contained during calibration
• 𝑚 = mass of filled vessel − mass of empty vessel
𝜌𝑤 = liquid density, in g/mL

𝜌𝑎 = air density, in g/mL

𝜌𝑏 = density of masses used during calibration of the balance assumed to be 8,0 g/mL
𝛾 = cubic thermal expansion coefficient of POVA, in °C-1 . For POVA, 𝛾 ≈ 0 (DKD−R 8−1).
𝑡𝑑 ≈ 𝑡𝑤 device temperature, in °C

5 [Link]
Density of water (Tanaka Formula)

𝑡𝑤 + 𝑎1 2 𝑡𝑤 + 𝑎2
𝜌𝑤 = 𝑎5 1−
𝑎3 𝑡𝑤 + 𝑎4
where,
𝑡𝑤 = water temperature in °C
𝑎1 = - 3.983 035 °C
𝑎2 = 301.797 °C
𝑎3 = 522 528.9 °C2
𝑎4 = 69.348 81 °C
𝑎5 = 0.999 974 95 g/mL

6 [Link]
Air Density (Spieweck’s formula)
𝑘1 𝑝𝑎 + 𝜑 𝑘2 𝑡𝑎 + 𝑘3
𝜌𝑎 =
𝑡𝑎 + 273.15
𝑡𝑎 = ambient temperature, in °C
𝑝𝑎 = ambient pressure, in hPa
𝜑 = relative air humidity, in %
𝑘1 = 3.4844 x 10−4 ºC/hPa
𝑘2 = −2.52 x 10−6 g/mL
𝑘3 = 2.0582 x 10−5 ºC
Spieweck’s equation is only applicable if: pressure is between 940 hPa and 1080 hPa;
temperature is between 18 ºC and 30 ºC; humidity less than 80 %

7 [Link]
Sources of Uncertainty for Volumetric Instrument

8 [Link]
Standard Uncertainty from each source
Model equation of the standard
measurement uncertainty:

9 [Link]
Standard Uncertainty from each source
1. Mass
a.1. Uncertainty of Balance w/ load

Half width estimate

Standard unc. 𝑈𝑏𝑎𝑙
𝑢 𝑏𝑎𝑙 = ,𝑘=2
𝑘
divisor
Example: Test point is 1 mL
𝑈𝑏𝑎𝑙 = 0.068 mg

0.068
𝑢 𝑏𝑎𝑙 = = 0.034 mg
2

a.2. Uncertainty of Balance w/out load

0.068
𝑢 𝑏𝑎𝑙 = = 0.034 mg
2

10 [Link]
Standard Uncertainty from each source
1. Mass

b.1. Resolution w/ load

𝑟𝑒𝑠
𝑢 𝑟𝑒𝑠 = ,k= 3
𝑘

Example: Test point is 1 mL; res. of balance = 0.01 mg

0.01/2
𝑢 𝑟𝑒𝑠 = = 0.002887 mg
3

b.2. Resolution w/out load

0.01/2
𝑢 𝑟𝑒𝑠 = = 0.002887 mg
3

11 [Link]
Standard Uncertainty from each source
1. Mass

c. Temperature Drift

𝑡𝑑
𝑢 𝑡𝑑 = ,k= 3
𝑘

𝑡𝑑 = sample mass x temp. coefficient of sensitivity x ∆𝑡𝑎

temp. coefficient of sensitivity = from balance specs.
∆𝑡𝑎 = maximum change of ambient temp. in the lab during
calibration

Ex: test point is 1mL; average mass of water obtained = 0.9989 mg

∆𝑡𝑎 = 0.5 °C ; temp. coefficient of sensitivity = 1 ppm/ °C
𝑡𝑑 = 0.9989 x 0.000001 x 0.5 = 4.9945E-07 mg

4.9945E−07
𝑢 𝑡𝑑 = = 2.88358𝐸 − 07 𝑚𝑔
3

12 [Link]
Standard Uncertainty from each source
1. Mass Calculation for Evaporation Loss (𝑒𝑣𝑎𝑝.):

d. Evaporation Loss 1. Perform 10 measurement cycles (𝑚1 𝑡𝑜 𝑚10 ).

Measure the time 𝑡𝑚 taken to complete 10 cycles.
𝑒𝑣𝑎𝑝.
𝑢 𝑒𝑣𝑎𝑝. = ,k= 3 Note: 1 cycle = obtaining the reference liquid to dispensing it in the
𝑘
receiving vessel.
Example: 2. After recording 𝑚10 , do not remove the receiving vessel from
weighing pan of the balance. Wait for time 𝑡𝑚 and record the
𝑒𝑣𝑎𝑝. = (𝑚10−𝑚11)ൗ10 = 0.010 uL new reading as 𝑚11 in the display of the balance.
3. Calculate for the evaporation loss per cycle (𝑒𝑣𝑎𝑝.), (𝑚10−𝑚11)ൗ10
0.010 uL
𝑢 𝑒𝑣𝑎𝑝. = = 0.005774 uL
3

13 [Link]
Standard Uncertainty from each source
2. Density of mass pieces
Assumed Uncertainty
Material
The value presented in the calibration certificate of the Density (k=2)
Platinum 21400 kg.m-3 150 kg.m-3
set of masses, or of the analytical balance can be Nickel silver 8600 kg.m-3 170 kg.m-3
Brass 8400 kg.m-3 170 kg.m-3
used. Alternatively the uncertainties corresponding to
Stainless Steel 7950 kg.m-3 140 kg.m-3
the used weight class according to OIML R 111-1 can
Table B.7 Method F2 - List of alloys
be used (𝑘 = 2). most commonly used for weights
with their expanded uncertainties
𝑈𝑏
𝑢 𝜌𝑏 =
𝑘
0.14 𝑔/𝑚𝐿
𝑢 𝜌𝑏 = = 0.07 𝑔/𝑚𝐿
2

14 [Link]
Standard Uncertainty from each source
3. Water Temperature

2 1/2
𝑈(𝑡𝑤 )
𝑢 𝑡𝑤 = + 𝑢2 𝛿𝑡𝑤
𝑘

• 𝑈(𝑡𝑤 ) & 𝑘 - found in the calibration certificate

• δ𝑡𝑤 - estimate of the uncertainty caused by possible drift and ageing of the temperature
measuring system after its calibration.

Ex: 𝑈(𝑡𝑤 )= 0.11 °C ; 𝑘=2 ; 𝑡𝑤 = 0

0.11
𝑢 𝑡𝑤 = = 0.55 °C
2

15 [Link]
Standard Uncertainty from each source
4. Water Density

Estimated uncertainty is 10 × 10−6 g/mL. For the formula used, 𝑘 = 2.

10 × 10−6
𝑢 𝜌𝑊 = = 5.00𝐸 − 06 g/mL
2

5. Air Density

Estimated uncertainty is 5 × 10−7 g/mL. For the formula used, 𝑘 = 3.

5×10−7
𝑢 𝜌𝑎 = = 2.89𝐸 − 07 g/mL
3

16 [Link]
Standard Uncertainty from each source
6. Air Temperature

2 1/2
𝑈 𝑡𝑎
𝑢 𝑡𝑎 = + 𝑢2 𝛿𝑡𝑎
𝑘

• 𝑈 𝑡𝑎 & 𝑘 - found in the calibration certificate

• δ𝑡𝑎 - estimate of the uncertainty caused by possible drift and ageing of the temperature
measuring system after its calibration.

• Ex: 𝑈 𝑡𝑎 = 0.56 °𝐶 ; 𝑘=2 ; δ𝑡𝑎 = 0.

0.56
𝑢 𝑡𝑎 = = 0.28 °𝐶
2

17 [Link]
Standard Uncertainty from each source
7. Air Pressure

2 1/2
𝑈 𝑝𝑎
𝑢 𝑝𝑎 = + 𝑢2 𝛿𝑝𝑎
𝑘

• 𝑈 𝑝𝑎 & 𝑘- found in the calibration certificate

• 𝛿𝑝𝑎 - estimate of the uncertainty caused by possible drift and ageing of the pressure measuring
system after its calibration.

• Ex: 𝑈 𝑝𝑎 = 0.05 hPa ; 𝑘=2 ; 𝛿𝑝𝑎 = 0

0.05
𝑢 𝑡𝑎 = = 0.025 ℎ𝑃𝑎
2

18 [Link]
Standard Uncertainty from each source
8. Air Humidity

2 1/2
𝑈 𝜑
𝑢 𝜑 = + 𝑢2 𝛿𝜑
𝑘

• 𝑈 𝜑 & 𝑘- found in the calibration certificate

• 𝛿𝜑 - estimate of the uncertainty caused by possible drift and ageing of the pressure measuring
system after its calibration.

• Ex: 𝑈 𝜑 = 1.6 % ; 𝑘=2 ; 𝛿𝜑 = 0

1.6
𝑢 𝑡𝑎 = = 0.8 %
2

19 [Link]
Standard Uncertainty from each source
9. System Effects

0.5
𝑢 𝑆𝐸∆𝑡 = = 0.2887 ℃
3

5
𝑢 𝑆𝐸∆𝜑 = = 2.8868 %
3

20
𝑢 𝑆𝐸∆𝑃𝑎 = = 8.1650 hPa
3

20 [Link]
Standard Uncertainty from each source
10. Process Handling Example:
These influences are from the following: Single channel; variable volume: 0.1 %
Actual volume: 99.7 uL
• Mechanical influences
• Operator-based influences 0.001 (99.7)
• Hand warmth 𝑢 𝛿𝑉𝑝ℎ = = 0.0576 uL
• Transport 3

% (𝑎𝑐𝑡𝑢𝑎𝑙 𝑣𝑜𝑙𝑢𝑚𝑒 𝑜𝑏𝑡𝑎𝑖𝑛𝑒𝑑)

𝑢 𝛿𝑉𝑝ℎ =
3

21 [Link]
Standard Uncertainty from each source
11. Repeatability

by excel function:
OR stdev (range of cells
containing the data)

22 [Link]
Sensitivity Coefficients
Sensitivity Coefficients are multipliers used to convert each uncertainty component to
obtain similar unit of measurement in the uncertainty budget. These conversion factors are
partial derivatives describing how the output estimate varies with the changes in the values
of the input estimates. The values of sensitivity coefficient are determined with the
following approximations:

1 1
≈ ; 𝜌𝑏 − 𝜌𝑎 ≈ 𝜌𝑏 ; 𝜌𝑏 − 𝜌𝑤 ≈ 𝜌𝑏 ;
𝜌𝑤 −𝜌𝑎 𝜌𝑤

1 − 𝛾 𝑡𝐷 − 25 ℃ ≈ 1

23 [Link]
Sensitivity Coefficients
Symbols
𝑚ഥ = average mass of the water obtained
𝜌𝑏 = density of mass pieces
𝜌ҧ𝑤 = average water density
𝑡𝑤 ҧ = average water temperature
𝜌ҧ𝑎 = average air density
𝑡𝑎ҧ = average ambient temperature of air
𝜑ത = average relative humidity
𝑝𝑎ҧ = average ambient pressure
𝑉ഥ𝑜 = average actual volume of water obtained
𝑎1 𝑡𝑜 𝑎4 = constants found in Tanaka equation
𝑘1 𝑡𝑜 𝑘3 = constants found in Speiweck’s equation

24 [Link]
Sensitivity Coefficients
1. The sensitivity coefficient 𝑐𝑚 related to the balance reading m:
ഥ
𝑚 𝜌𝑏 −ഥ
𝜌𝑎
𝑉𝑜 = · ഥ𝑤 −ഥ
· 1 − 𝛾 𝑡𝑤ഥ − 25 ℃
𝜌𝑏 𝜌 𝜌𝑎

ഥ0
𝜕𝑉
𝑐𝑚 = ഥ
=?
𝜕𝑚
Step 1: separate variable that are treated as constants:
ഥ0
𝜕𝑉
𝑐𝑚 = ഥ = 𝜌ഥ𝜌𝑏−ഥ 𝜌𝑎
· 1 − 𝛾 𝑡𝑤ഥ − 25 ℃ 𝑥
1
𝑥
𝜕
ഥ
(𝑚)
ഥ
𝜕𝑚 𝑤 −ഥ
𝜌𝑎 𝜌𝑏 𝜕𝑚

Step: differentiate m in terms of 𝑚 in terms of 𝑚 :

ഥ0
𝜕𝑉 𝜌𝑏 −ഥ
𝜌𝑎 1 𝜕
𝑐𝑚 = ഥ
= ഥ𝑤 −ഥ
· 1 − 𝛾 𝑡𝑤ഥ − 25 ℃ 𝑥 𝑥 ഥ
(𝑚)
ഥ
𝜕𝑚 𝜌 𝜌𝑎 𝜌𝑏 𝜕𝑚
1 1
Step 3: Recall that 𝛾= 0 for piston pipette, 𝜌𝑏 − 𝜌𝑎 ≈ 𝜌𝑏 and ≈ . Thus,
ഥ𝑤 −ഥ
𝜌 𝜌𝑎 ഥ𝑤
𝜌
ഥ0
𝜕𝑉 𝜌𝑏 1
𝑐𝑚 = ഥ
= ·0 𝑥 𝑥1
𝜕𝑚 ഥ𝑤
𝜌 𝜌𝑏

𝟏
𝒄𝒎 =
ഥ𝒘
𝝆

25 [Link]
Sensitivity Coefficients

2. Temperature of reference liquid

ҧ + 𝑎1 2 ҧ + 𝑎2 + 𝑡𝑤
ҧ + 𝑎4 𝑡𝑤ҧ + 𝑎1 2 ҧ + 𝑎4 ҧ + 𝑎2 𝑡𝑤
ҧ + 𝑎1
𝑚
ഥ 𝑎5 − 𝑡𝑤 𝑡𝑤 + 2 𝑡𝑤 𝑡𝑤
𝑐𝑡𝑤 = 2 × ×
𝜌ҧ𝑤 𝑎3 ҧ + 𝑎4
𝑡𝑤 2

3. Air Pressure

𝑚
ഥ 𝑘1
𝑐𝑝𝑎 = .
𝜌ҧ𝑤 2 𝑡𝑎ҧ + 273.15

26 [Link]
Sensitivity Coefficients

4. Air Humidity

𝑚
ഥ 𝑘2 𝑡𝑎ҧ + 𝑘3
𝑐𝜑 = .
𝜌ҧ𝑤 2 𝑡𝑎ҧ + 273.15

5. Temperature of air

𝑚
ഥ 𝜑ത 273.15 𝑘2 − 𝑘3 − 𝑘1 𝑝𝑎ҧ
𝑐𝑡𝑎 = 2𝑥
𝜌ҧ𝑤 𝑡𝑎ҧ + 273.15 2

27 [Link]
Sensitivity Coefficients

6. Water Density

𝑚
ഥ
𝑐𝜌𝑤 = −
𝜌ҧ𝑤 2

7. Air density

− 𝜌ҧ𝑤 + 𝜌𝑏
𝑐𝜌𝑎 = 𝑚
ഥ 𝑥
𝜌𝑏 𝜌ҧ𝑤 2

8. Density of Mass Pieces

𝑚ഥ 𝜌ҧ𝑎
𝑐𝜌𝑏 =
𝜌ҧ𝑤 𝜌𝑏 2

28 [Link]
Sensitivity Coefficients

9. Process Handling

𝜕𝑉𝑜
𝑐𝛿𝑉𝑝ℎ = =1
𝜕𝛿𝑉𝑝ℎ

10. Repeatability

𝜕𝑉𝑜
𝑐𝛿𝑉𝑟𝑒𝑝 = =1
𝜕𝛿𝑉𝑟𝑒𝑝

29 [Link]
Sensitivity Coefficients
11. System Effects

Uncertainty due to system effects is estimated for the calibration of piston pipettes in which the
influences depend, in particular, on the size of the air cushion and on the lifting height in the
pipette tip.

Sensitivity coefficient related to temperature Difference of water-pipette-air

From experimental data of previous studies, pipettes with nominal volumes of about 1 000
µl, 500 µl, 100 µl have sensitivity coefficients of approximately 0.22 % change due to 1 K
temperature difference between water and pipette/air. Other nominal volumes have
different coefficients, but the deviation is small.

𝑐𝑆𝐸∆𝑡 = 0.0022 (𝑉ഥ𝑜 )

30 [Link]
Sensitivity Coefficients
11. System Effects

Sensitivity coefficient related to air moisture/ relative humidity

Latest studies indicate that pipettes with nominal volumes of 1 000 µl, 500 µl, 100 µl have
sensitivity coefficients of about 0.07 % in accuracy of the aspirated volume per 10 % RH
change of moisture.

0.0007
𝑐𝑆𝐸∆𝜑 = (𝑉ഥ𝑜 )
10 %

31 [Link]
Sensitivity Coefficients
11. System Effects

Sensitivity coefficient related to atmospheric pressure

The sensitivity coefficient related to atmospheric pressure (𝑝𝑎 ) can be computed by taking
the partial derivative of the equation of volume difference resulting from changes in
atmospheric pressure during calibration from location x1 to x2.

𝜕𝑉 −𝑉𝑇 . 𝜌ത𝑤 . 𝑔 . ℎ𝑤
𝑐𝑆𝐸∆𝑃𝑎 = ≈
𝜕𝑝𝑎,𝑥2 ҧ 22
𝑝𝑎,𝑥
where,
𝑉𝑇 = volume of air cushion
𝑔 = acceleration due to gravity
ℎ𝑤 = lifting height of the test liquid
𝑝𝑎,𝑥2 = ambient pressure
32 [Link]
Sensitivity Coefficients
11. System Effects

Sensitivity coefficient related to atmospheric pressure

Table 2. Maximal values for liquid lift and air cushion (DKD-E 8-2)
Nominal Volume, µl Liquid lift in the tip (ℎ𝑤 ), mm Air cushion (𝑉𝑇 ), µl

> 5 000 to 10 000 150 20000

> 1 000 to 5 000 130 10000
> 300 to 1 000 55 3000
> 100 to 300 45 1000
> 10 to 100 30 500
1 to 10 19 100

33 [Link]
Uncertainty Contribution of each source
Uncertainty Contribution = standard uncertainty x sensitivity coefficient

𝑢 𝑉𝑜 𝑥𝑖 = 𝑢(𝑥𝑖 ) . 𝐶𝑥𝑖

Ex: Uncertainty Contribution of repeatability

𝑢 𝛿𝑉𝑟𝑒𝑝 = 0.00109276 ml & 𝐶𝛿𝑉𝑟𝑒𝑝 = 1

𝒖 𝑽𝒐 𝜹𝑽𝒓𝒆𝒑 = 0.00109276 (1) = 0.00109276 mL

34 [Link]
Combined Standard Uncertainty

2
𝑢𝑐 = ෍ 𝑢 𝑉𝑜 𝑥𝑖
𝑖

• All the uncertainty contributions are sum up by Summation of Quadrature or Root Sum of the Squares to
obtain the combined standard uncertainty. It can be computed using the equation above or by using the
function “sqrt(SUMSQ(range of cells containing the uncertainty contributions))” in excel,

35 [Link]
Coverage Factor, k

• a coverage factor of k = 2 at approximately 95% confidence

interval corresponds to 10 observations. Thus, it is
recommended to perform 10 measurements to obtain k = 2.

36 [Link]
Expanded Uncertainty

U is obtained by multiplying 𝒖𝒄 with the coverage factor (k).

𝑼 = 𝒖𝒄 𝑥 𝒌

37 [Link]
References

• DKD-R 8-1:2011 Calibration of piston-operated pipettes with air

cushion
• DKD-E 8-2:2013 Analysis of influencing parameters on calibration
of piston-operated pipettes with air cushions
• ISO 8655:2002 Piston-operated volumetric apparatus
• JCGM 100:2008 Evaluation of measurement data — Guide to the
expression of uncertainty in measurement

38 [Link]
For more information about our services, please visit our
website [Link], or send us an email to
info@[Link] or nmlphil@[Link]

39 [Link]
Coverage Factor, 𝒌

Effective Degrees of Freedom using the Welch-Satterthwaite

formula, 𝒗𝒆𝒇𝒇 :

40 [Link]
Coverage Factor, 𝒌
Method 1
1. Open a MS Excel workbook.
2. Select a cell and type “TINV”.
3. Enter your probability value and your degrees of freedom.
4. Hit Enter and find your coverage factor.

• TINV(probability, effective degrees of freedom)

• To find the probability:

Probability = (1-α) = (1-0.95) = 0.05

41 [Link]
Coverage Factor, 𝒌
1. Refer to the Student’s T-table.
2. Find the column that matches
your confidence interval.
Ex: 𝒗𝒆𝒇𝒇 = 4;
95% probability
4. Find the row matching
your 𝒗𝒆𝒇𝒇 .
5. The value where your column and
row meet is your coverage
factor.

42 [Link]