Measurement Validation and Uncertainty of An Exper
Measurement Validation and Uncertainty of An Exper
Measurement Validation and Uncertainty of An Exper
Instituto Nacional de Técnica Aeroespacial (INTA), Centro de Metrología y Calibración, Ctra. de Ajalvir, km 4,
28850 Torrejón de Ardoz, Spain; [email protected]
Abstract: An experimental procedure for characterizing the size-of-source effect (SSE) is proposed.
Such an effect is the cause of one of the main influence variables generating uncertainty in the meas-
urement, both in calibration and use, of direct reading radiation thermometers (RT). The procedure
and uncertainty calculation described in the paper are aligned in terms of metrological traceability,
with the requirements generally imposed to ensure the accuracy of measurements in industry and
science. Results of application and validation of this particular procedure with equipment, includ-
ing black body (BB) sources normally used in radiation thermometry calibration laboratories in the
industrial field, are shown.
order to facilitate the calculation of corrections to the RT reading, between the calibration
diameter dc and other different d, (3) is used to define the quantity:
𝑆𝑇,𝑑 − 𝑆𝑇,𝑑𝑐
𝜎 ∗ (𝑑, 𝑑𝑐 ) = 𝜎(𝑑) − 𝜎(𝑑𝑐 ) = (4)
𝑆𝑇,𝑑0
In first approximation, it can be assumed σ(d) 1 and, with (4), it is easy to see that:
𝜉(𝑑) ≅ 𝜉(𝑑𝑐 )(1 + 𝜎 ∗ (𝑑, 𝑑𝑐 )) (5)
For an input signal S, direct reading RT shows a temperature value I (display indica-
tion), implicitly dependent on temperature and diameter (area in general) of the source, I
(T, d). Typically, this kind of RT is provided with an emissivity function εinstr selectable
between 0 and 1. This parameter is called instrumental emissivity adjustment and facili-
tate the use of RT to measure temperature of surfaces different to perfect black bodies. In
general, I also depends on detector temperature Td and ambient temperature Tw [9]. In
normal laboratory ambient conditions, it can be assumed Td Tw when RT is not too close
to the BB aperture at high temperatures. Provided that there are no additional correction
factors aside from SSE, I can be obtained from:
𝜀𝑎 (𝜀𝑖𝑛𝑠𝑡𝑟 − 𝜀𝑎 )
𝑆𝑠ℎ (𝐼(𝑇, 𝑑)) = 𝜉(𝑑) 𝑆𝑠ℎ (𝑇) + 𝑆𝑠ℎ (𝑇𝑤 ) (6)
𝜀𝑖𝑛𝑠𝑡𝑟 𝜀𝑖𝑛𝑠𝑡𝑟
Both T and d are considered the main variables, while εa, εinstr and Tw are fixed param-
eters. Typically, laboratory conditions allow simplify (6), by selecting εinstr = 1, and by the
use of a cavity BB with εa 1 at a temperature T much larger than Tw. Then, the second
summand in (6) can be neglected and the correction I, when RT is used to measure the
temperature of a source with diameter different from that used in calibration dc, can be
obtained calculating Ssh(I(T, dc + d))/Ssh(I(T, dc)). Considering, by definition, I(T, dc + d) =
I + I and d = dc + d, by developing at the first order, we obtain:
𝑆𝑠ℎ (𝐼)
Δ𝐼 = 𝜎 ∗ (𝑑, 𝑑𝑐 ) (7)
(𝑑𝑆𝑠ℎ ⁄𝑑𝑇 )(𝐼)
With (1), we can finally calculate the correction to RT temperature reading I, as a
function of source diameters d:
𝜆𝑥 𝐼 2 −𝑐2
𝐶𝑆𝑆𝐸 = Δ𝐼 = 𝜎 ∗ (𝑑, 𝑑𝑐 ) (1 − 𝑒 𝜆𝑥𝐼 ) (8)
𝑐2
Figure 1 shows schematically the basic principles of SSE measurement.
Stops delimiting BB
Radiation reaching
variable apertures
the detector due SSE
ea d0 Field stop
d
RT detector
Contact thermometers
for measure and
T dt
control BB
temperature
Nominal
550 C
target 545 C
RT objective
FOV lens
Figure 1. Basic configuration for SSE measurement of a RT, with a BB cavity source of variable ap-
erture. Dashed gray arrows represent rays reaching the RT detector due SSE. Solid red and black
arrows represent rays reaching the detector, within the theoretical field of view (FOV).
In (8), λx depends on RT spectral responsivity (optics, filters and detector). Its value
can be calculated by means of direct calibration of RT against reference BB, using (1) for
Sensors 2022, 22, 8284 5 of 17
fitting the parameters A, B and C, provided that RT internal signal, linear with the radi-
ance, may be accessible [10, 11]. Nevertheless, in industrial direct-reading RT this is not
possible in general. For SSE characterization and uncertainty evaluation of this type of
instruments, it is still possible to assume rectangular spectral response in the waveband
λ = (λ2 - λ1) [12]. Then, it can be proved that A = λ0(1–6(ζ/λ0)2) and B = (c2/2)(ζ/λ0)2, being
λ0 = (λ1 + λ2)/2 and ζ = λ/√12.
For non-uniform radiation sources, the mathematical model for calculating SSE cor-
rections is quite complex and difficult to solve in general. If the element delimiting aper-
ture (as in Figure 1) is very close to BB, cooling the stops could not be possible and, de-
pending on RT, radiation emitted by these elements, may not be negligible. The set formed
with the BB and the element defining aperture defines an effective (or apparent) diameter
deff considered as that of a uniform source producing the same signal in the RT [13]. The
value deff depends on the particular RT and also on the temperature gradients existing in
the piece defining the aperture, including the front plate of the source.
2.2.1. Reference BB
The BB used in this experimental procedure is a source designed and manufactured
by National Physical Laboratory NPL, UK, for INTA in 1993, based on a commercial
GRANT calibration bath, provided with an oxidized stainless steel cylindrical cavity, with
surface intrinsic emissivity ε = 0.85, for the spectral band between 0.9 μm and 14 μm. With
length 347 mm and diameter 77 mm, the bottom cavity is grooved (as seen in Figure 1) in
order to increase internal reflections and therefore εa. The cavity is immersed in a silicone
oil isothermal bath with a PID temperature controller, enabling stability and uniformity
better than ±0.05 °C around the cavity, in the temperature range between 30 °C and 180
°C. Figure 2 shows an opening at the front of the tank, where the cavity is attached, allow-
ing radiation towards the exterior.
Sensors 2022, 22, 8284 6 of 17
PID controller
PRT cavity PRT bath
Apertures
support piece
415 mm
Heater
77 mm
d0=70
mm
Cavity Aperture
347 mm
Silicone oil
Stirrer
(a) (b)
Figure 2. (a) BB used in this research as reference radiation source for SSE characterization; (b) De-
tailed scheme of the system, showing the bath, reference thermometers, cavity and support piece
for the apertures. Red arrows represent rays emitted and reflected by the BB internal surfaces.
the Monte Carlo method, (εa(1 − εa)/N)1/2 1.4 · 10−5, N = 107 being the number of trials
computed.
190 0.9985
8 microns 14 microns
170 (I) 0.9984
effective emissivity
temperature [ºC]
150
(II) 0.9983
130
0.9982
110
0.9981
90 (III)
70 0.9980
0 100 200 300 400 0 10 20 30
length [mm] radial distance from the bottom center [mm]
(a) (b)
Figure 3. (a) Longitudinal temperature gradient measured at the BB cavity; (b) Radial profiles for
the effective emissivity in 8 μm and 14 μm, as a function of the distance from the center.
The estimation of εa and thus the radial uniformity of the BB radiance temperature,
was carried out averaging the values showed in Figure 3b. The maximum variation was
calculated as half of the difference between the maximum at 14 μm and the minimum at
8 μm. The results were εa ± εa = 0.9982 ± 1.4 · 10−4, εa, in terms of radiance temperature,
represents 0.02 °C at 8 μm and 0.03 °C at 14 μm, when BB temperature is 180 °C. Taking
into account the expected uncertainties in the SSE corrections, the estimated temperature
radiance uniformity through the 70 mm aperture is considered very acceptable.
Apertures
supporting
piece
Brass aperture
piece
5 ºC
Recirculating bath
(a) (b)
Sensors 2022, 22, 8284 8 of 17
(c) (d)
Figure 4. (a) Interchangeable brass pieces defining the variable apertures; (b) Scheme of the aper-
tures cooling system. Orange arrows indicate the fitting of the brass pieces on the structure; (c)
Structure for positioning the apertures in front of BB; (d) Cooling recirculating bath.
With the system described, the apertures can be cooled up to 5 °C. It has been ob-
served that the time required to stabilize the temperature of the pieces defining apertures,
after each operation of piece changing is about 10 min. Then, to avoid errors due to a
possible drift of BB temperature (the effect of BB stability), continuous recording with the
cavity PRT is needed. The RT readings for each aperture, have to be corrected and stand-
ardized to the temperature corresponding to the first aperture. We will come back to this
in the next section.
2.4. Measurement Process, Model for Analysis of the Results and Uncertainty Calculation
Once the source and aperture temperatures are stabilized, a representative number
of RT readings I is taken in order to obtain average and standard deviation (measure of
dispersion or repeatability). The reproducibility of the procedure is evaluated by taking a
second series of measurements, refocusing and realigning the RT on the aperture of minor
diameter (d1), always searching for the maximum signal. It is advisable to select d1 large
enough (and of course > dt) because small errors in target centering can lead to dispersion
in the results and thus, to an increasing of uncertainty in SSE characterization for small
sources.
With the pairs obtained experimentally {di, Ii}, being dt < di ≤ d0, we calculate σi*(di, dc)
with (1) and (4), being ST,d = Ssh(I), defining C = 1, and with λx(A, B) function of the RT
bandwidth. The corresponding corrections between arbitrary diameters dy and dz are cal-
culated with σ*(dy, dz) = σ*(dy, dc) - σ*(dz, dc). For data fitting, we have proposed a reference
Sensors 2022, 22, 8284 9 of 17
function σSSE(d), that represents adequately the general behavior observed in most com-
mercial direct-reading RT. The fitting function must fulfil σSSE(dc) = 0 and σSSE(d) → con-
stant, when d → . Taking this in mind, we have opted by a function dependent of three
free parameters (a, b1, b2):
2
𝜎𝑆𝑆𝐸 (𝑑) = 𝑎(1 − 𝑒 𝑏1(𝑑𝑐 −𝑑)+𝑏2 (𝑑𝑐 −𝑑) ) (9)
If, as it is expected, d0 is large enough, in general it holds a σ*(d0, dc) and the calcu-
lation of b1 and b2 reduces to a least squares fitting of the set {di, log(1 - σi*/a)} to the function
(linear in the parameters): b1(dc - d) + b2(dc - d)2.
The uncertainty calculation of the RT reading I, due to all the significant influence
quantities or variables affecting the temperature measurement with this kind of instru-
ments [7], is a complex task that exceeds the general objectives and aims of this paper. In
this work, we only consider the uncertainty u(σ*) and how it propagates to the uncertainty
of the correction I, that has to be applied to the indication I. As usual, we follow the
methodology given in GUM guide [16], for calculation, propagation and expression of
standard u and expanded U = 2u (or for k = 2) uncertainty.
The measurement model function for σ* is given in Equation (4). It depends on five
influence quantities {xk}k=1,…,5 : diameters d and dc, RT reading I for source diameter d,
reading Ic for the calibration diameter dc and I0 for the maximum diameter d0. The mathe-
matical model bases on the validity of (5) independently of d0, so this quantity is not an
influence variable for the uncertainty. The five characteristic variables of the model are
independent (no correlation or negligible) and therefore, the standard uncertainty can be
written as:
5
∗)
𝜕𝜎 ∗ 2 2
𝑢(𝜎 = √∑ ( ) 𝑢 (𝑥𝑘 ) (10)
𝜕𝑥𝑘
𝑘=1
With some algebra, it can be seen that the contribution of readings is:
𝜕𝜎 ∗ 2 2 𝜕𝜎 ∗ 2 2 𝜕𝜎 ∗ 2 2
( ) 𝑢 (𝐼) + ( ) 𝑢 (𝐼𝑐 ) + ( ) 𝑢 (𝐼0 )
𝜕𝐼 𝜕𝐼𝑐 𝜕𝐼0
2 (11)
𝑐2
= ((1 + 𝑆𝑠ℎ (𝐼0 )) ) (𝑢 2 (𝐼)
+ 𝑢2 (𝐼𝑐 ) + 𝜎 ∗ 2 𝑢2 (𝐼0 ))
𝜆𝑥 𝐼0 2
certainty component, is defined by u2(Ii) = │Cimed - Cical│/√3. The reproducibility of the pro-
cedure is also taken into account. We consider the difference between two measurements
series, the second one corresponding to a new RT realigning and refocusing, as was ex-
plained in Section 2.4. Thus, the contribution is defined by u3(Ii) = │Iiseries1 - Iiseries2│/√12, where
explicitly we assume that the current indication Ii is the average between that of the two
series. Combining the three components, we have finally:
2 2
𝑟 𝑠𝑖 2 |𝐶𝑖𝑚𝑒𝑑 − 𝐶𝑖𝑐𝑎𝑙 | |𝐼𝑖𝑠𝑒𝑟𝑖𝑒𝑠1 − 𝐼𝑖𝑠𝑒𝑟𝑖𝑒𝑠2 )|
𝑢2 (𝐼𝑖 ) = (max { ; }) + ( ) +( ) (13)
√12 √𝑛 √3 √12
The uncertainty of the correction between two arbitrary diameters (within the range
between d1 and d0) is calculated with (8). For that purpose, we consider it necessary to add
another uncertainty factor to (10) uextra, which quantifies the quality of the fit. It is common,
being conservative, to take the maximum residual considered as the absolute limit of a
rectangular distribution function, i.e., uextra = max{│σi* - σSSE(di)│}/√3.
3. Results
The procedure described in this work has been applied to typical direct-reading RT
used commonly in industrial applications, either in a laboratory environment or in indus-
trial facilities. In general, these thermometers are equipped with thermal sensors (of the
thermopile type) and work in the bandwidth 8 μm to 14 μm for temperature measuring
in the range between −50 °C and 1000 °C. RTs of this type are routinely used and/or cali-
brated and certified in calibration laboratories and also used in industrial applications.
They are used not only for temperature measuring but also as transfer standards (or com-
parators) in calibration of BB sources (of the cavity type or extended area) against refer-
ence or primary sources. The RTs we are selected are shown in Figure 5.
(a) (b)
(c) (d)
Figure 5. Commercial direct reading RTs used in this research. (a) Model IRCON UX40P; (b) Model
FLUKE 574CF; (c) Model LAND C300; (d) C300 during measurements.
The RTs were mounted into opto-mechanical mounts in front of the BB aperture, as
shown in Figure 5d. BB temperature was set to 180 °C and the cooling system was adjusted
to maintain the brass pieces defining apertures at 5 °C, avoiding water vapor condensa-
tion with the aid of a permanent flow of dry N 2 on the pieces. Considering such source
and aperture temperatures, the influence on the RTs of thermal radiation from the outer
surface of the BB aperture is not very significant. This means that source effective diame-
ters deff (depending on the particular RT and aperture used in each case) can be considered
as that of the aperture d.
The results for σ*, expressed as a percentage of the maximum reading (with d0 = 60
mm), are shown in Figure 6, considering calibration diameter dc = 30 mm. Figure 7 repre-
sents the correction, as a function of source temperature, to be applied to RT readings
between diameter 50 mm and diameter 30 mm. In terms of uncertainty, we have assumed
for U(di) a value of 0.5 mm (expanded uncertainty for k = 2). The error bars in Figure 6,
represent the expanded uncertainty U(σ*), obtained applying the model described above.
1.0
0.0
SSE s* [%]
-1.0
C300
-2.0 574CF
UX40P
-3.0
-4.0
0 10 20 30 40 50 60 70
Aperture diameter [mm]
Figure 6. Values for σ* expressed in %, as a function of aperture diameter, for the three RTs analyzed.
Sensors 2022, 22, 8284 12 of 17
0.0
-1.0
-2.0
CSSE [ºC]
-3.0
-4.0 C300
-5.0 574CF
-6.0 UX40P
-7.0
0 200 400 600 800 1000
Source temperature [ºC]
Figure 7. Corrections for SSE, to be applied to the three RT, as a function of temperature, between
source diameters 30 mm and 50 mm.
4. Discussion
The results obtained serve to illustrate the application of the procedure to various RT
models. It is evident from Figures 6 and 7 that SSE correction is very significant when it is
compared with calibration uncertainties. Considering the three models analyzed, the un-
certainties vary between 1.5 °C and 2.0 °C (expanded for k = 2) from 300 °C to 900 °C. As
mentioned above, it is clear that SSE can be a very significant and primary source of un-
certainty, either in calibration or use.
The lower uncertainty U(σ*) of model C300, when it compared with that of the others,
is mainly due to a lower contribution by effect of diameter indetermination. As can be
seen in (12), this influence factor in uncertainty strongly depends on the correction CSSE.
Due to its technical characteristics, model C300 is also used in calibration laboratories
(generally of the secondary type), acting principally as a comparator, for calibrating BB
sources of all types against primary BB. It has better quality optics and less (a priori) im-
perfections, resulting in SSE. However, this cannot be taken as a general conclusion. In the
course of calibration works, we have found that certain low-cost industrial RT, apparently
with a low or limited metrological level, perform better in terms of SSE than others con-
sidered to be of higher quality.
In order to validate the procedure developed in this work, we have proposed the use
of a high temperature BB source equipped with four apertures plates of diameters: 20 mm,
30 mm, 40 mm and 50 mm. Therefore, we have tried to prove if (within the uncertainty),
the calculated SSE corrections are compatible with RT readings for temperatures and
source different from that used to characterize this effect. To this end, we have used a BB
cavity source specified for the temperature range between 150 °C and 1100 °C. The cavity
is of the cylinder-conical type and is made of SiC (silicon carbide), a ceramic material with
a high intrinsic emissivity. With a length of 300 mm, diameter 50 mm and apex cone angle
120°, its effective emissivity εa was numerically calculated [18], resulting εa = 0.998 for the
bandwidth 8 μm to 14 μm. The cavity is installed in a three-zone furnace (model
LAND/CARBOLITE, LandCal P1200B) equipped with two standard type R thermocou-
ples (calibrated with expanded uncertainty U(t90) = 0.6 °C), for measuring cone apex tem-
perature and longitudinal gradient. The BB source P1200B and aperture plates are shown
in Figure 8.
Sensors 2022, 22, 8284 13 of 17
Figure 8. BB source and apertures plates used to validate the SSE characterization procedure.
To validate the procedure, we have used the model C300 due to its lower uncertainty.
This fact allows the calculation of a more representative compatibility index (En) for the
two series of values (calculated and measured). The quantity En measure the compatibility
and is the standard figure of merit in inter-comparison exercises in metrology [19]. It is
defined as:
|𝑣1 − 𝑣2 |
𝐸𝑛 = (14)
√𝑈 2 (𝑣1 ) + 𝑈 2 (𝑣2 )
where v1 and v2 are the values of the quantities to compare, U(v1) and U(v2) being the cor-
responding expanded uncertainties for a given coverage factor k (usually 2). In general,
provided En ≤ 1, the compatibility between the values is fulfilled.
The results for source temperatures 300 °C, 500 °C, 700 °C and 900 °C are shown in
Table 2. For each BB temperature and each pair of apertures, the correction CSSE (8) is
shown, together with its uncertainty USSE, and the difference RT between the RT readings
considering both apertures, also with its uncertainty URT. Such uncertainties were calcu-
lated considering specific influence factors, arising from the use of the high temperature
BB as the uniformity and stability of the radiance temperature, display resolution of the
C300 thermometer, reproducibility of the measurement procedure, etc.
Table 2. Results for the compatibility of the procedure for calculating SSE corrections, with the RT
model C300 and the BB of Figure 8.
Apertures
CSSE, °C USSE, °C RT, °C URT, °C En
(Diameter), mm
300 °C
30→20 −1.0 0.2 −1.1 0.1
30→40 0.4 0.1 0.4 0.4 0.1
30→50 0.6 0.1 0.6 0.1
500 °C
30→20 −1.8 0.3 −1.9 0.2
30→40 0.7 0.2 0.6 0.7 0.1
30→50 1.0 0.2 0.8 0.3
700 °C
30→20 −2.5 0.4 −2.4 0.1
30→40 1.0 0.4 0.9 1.1 0.1
30→50 1.5 0.4 1.3 0.2
900 °C
Sensors 2022, 22, 8284 14 of 17
With the mentioned criterion En ≤ 1, all the measurements are compatible within the
calculated uncertainty and under experimental conditions which can be reproduced in an
industrial calibration laboratory framework. As stated above, the model assumes that ra-
diation coming from the surface exterior to the apertures is negligible, which is possible
when the temperature of the brass pieces is very low compared to that of the source. At 5
°C and 180 °C respectively, we found the described set-up suitable for this purpose.
T << TBB
Piece delimiting aperture, at
deff temperature TE
dE
Then, deff depends on the variables dE, d, TBB, TE and also on the specific characteristics
of the RT in regard to its SSE corrections, evaluated as explained above.
The general model for extended and non-uniform sources, with arbitrary tempera-
ture gradients is complex [4]. However, with the previous hypothesis, it can be proved
that deff can be calculated as implicit solution of an equation that can be easily experimen-
tally implemented, being:
∗ ∗
𝑒 𝑐2/(𝜆𝑥⋅𝐼𝐵𝐵) − 1 ∗ ∗
𝜎𝑎,𝑏1 ,𝑏2
(𝑑𝑒𝑓𝑓 ) = 𝜎𝑎,𝑏1 ,𝑏2
(𝑑) + [ ] (𝜎𝑎,𝑏1 ,𝑏2
(𝑑𝐸 ) − 𝜎𝑎,𝑏1 ,𝑏2
(𝑑)) (15)
𝑒 𝑐2/(𝜆𝑥⋅𝐼𝐸 ) − 1
where (for εinstr = 1) IBB is the RT indication when focused on the BB aperture center, IE is a
spatial average when focused on the outer surface (assuming some temperature non-uni-
formity) and σ*, as we saw above, explicitly depends on diameter and parameters a, b1
and b2 characterizing the SSE of the particular RT. With σ* obtained under standard con-
ditions (with a uniform BB source and no thermal radiation from the surroundings), deff
can be calculated from (15), either analytically or numerically.
Sensors 2022, 22, 8284 15 of 17
One of the circumstances where correction for SSE is essential occurs in the course of
inter-laboratory comparison exercises, where the accredited calibration laboratories have
to periodically participate. This type of activity is essential, within the requirements of
standard 17025 [3], for assuring the validity of the measurement results. In order to com-
pare calibrations of RT performed in several laboratories, the results have to be standard-
ized to equal source diameter. Generally, it is the diameter used by the laboratory acting
as reference. This laboratory usually characterizes the RT SSE at standard conditions and
calculates corrections for each temperature, as a function of the diameters provided by the
participants, who rarely give effective apertures. It would be a significant improvement if
the laboratories would also report the readings when the RT is focused on certain points
of a well-defined region of the aperture surroundings, for example, for various diameters
greater than that of the aperture. The reference laboratory would then have information
to assign effective diameters to each participant, thus lowering the uncertainty corre-
sponding to the size of source indetermination. As we have said, this influence factor
strongly contributes to the uncertainty, mainly at high temperatures.
5. Conclusions
The procedure described has proven to be valid for its application to commercial RTs,
allowing industrial calibration laboratories to have a useful tool to improve the quality
and applicability of the results included in the calibration certificates they issue. It only
requires having a sufficiently uniform and well characterized radiation source and a set
of interchangeable elements to define variable apertures. The main advantage of this pro-
posal is that it allows the use of low temperature BB (< 200 °C). Generally, BBs of this type
have larger apertures, and they are more uniform in radiance temperature than the BB
designed for higher temperatures. As stated above, the results improve if the elements
defining apertures can be cooled to minimize the effect of background radiation.
Making corrections to RT readings due to SSE is essential for calibrating BB when the
RT is used as a comparator and the reference and calibrated sources have different aper-
ture sizes. On the other hand, for applications where environmental conditions (or setup)
in calibration, differ from that in use, the implementation of the method is more compli-
cated or generates more uncertainty, since in general there no well-defined size of source.
The common approach is to search for measurement configurations where the source can
be considered infinite in size, avoiding focusing the RT on regions with strong tempera-
ture gradients.
As an example of the importance of making SSE corrections when calibration uncer-
tainties in radiation thermometry are taking into account, we can consider an accredited
calibration laboratory (say CLab1) with accredited CMC (Calibration and Measurement
Capabilities) of ±2 °C (k = 2) at 500 °C in the waveband (8 to 14) μm. Let CLab0 the refer-
ence laboratory (higher metrological level in the traceability chain from SI) from which
CLab1 obtain its metrological traceability in radiation thermometry. If CLab1 uses a RT to
calibrate his 50 mm BB source, but this RT has been calibrated by CLab0 with a standard
30 mm BB, then (from Table 2) a correction of 1 °C has to be applied for a correct use of
the calibration certificate given by CLab0. Then the correction cannot be considered neg-
ligible, being 50% of the CMC.
SSE in thermal imaging systems is generally insufficiently studied. Although in some
aspects it is a phenomenon similar to that of RTs, there are significant differences. In ad-
dition to the effect of radiation deviating from its theoretical path due to imperfections in
the optics, internal reflections, diffraction, etc., there is an essential contribution due to
thermal and electrical interference between neighboring pixels [20]. The increasing use of
equipment of this type in the industrial framework, due to the obvious advantage of ther-
mal image (in addition to the temperature measurements), makes it essential that the cal-
ibration laboratories responsible for ensuring metrological traceability of the measure-
ments have a particular methodology to evaluate SSE also in TI systems, especially that of
the quantitative type.
Sensors 2022, 22, 8284 16 of 17
In calibration work, TI are usually considered simply as RT with a target size calcu-
lated based on a certain selected pixel group in the focal plane array (FPA). It is necessary
to study more in-depth the relationship between the classical SSE (originally defined for
RT), and the specific parameters or figures of merit associated with TI, such as the point
spread function (PSF) [21, 22]. The problem, especially in the LWIR region, is to generate
a point thermal source, avoiding thermal radiation from the region around that source.
For classical qualitative TI, where the quality of thermal images is much more important
than the accuracy of the temperature values shown, standardized tests are used to deter-
mine the minimum resolvable temperature difference (MRTD) based on the spatial and
thermal frequency of a slit pattern. Such tests are generally considered pass/fail tests and
are not used for calibration or calculation of correction factors, as required by quantitative
applications. Since technological developments and industrial and scientific applications
are increasingly directed towards quantitative thermography, it will be necessary to de-
velop specific procedures that incorporate classic techniques typical of RT to the stand-
ardized procedures typical of qualitative TI.
Finally, the method described for the calculation of effective diameters, can be a use-
ful tool for the improvement of results, either in RT calibration or in inter-comparison
exercises. Although the model described is quite simple, it implies an improvement when
the background temperature of the BB cannot be completely neglected.
References
1. Bureau International des Poids et Mesures (BIPM). The International System of Units (SI), 9th ed.; Bureau International des Poids
et Mesures (BIPM): Sèvres, France, 2019.
2. BIPM; IEC; IFCC; ILAC; ISO; IUPAC; IUPAP; OIML; BIPM Joint Committee for Guides in Metrology. International Vocabulary
of Metrology—Basic and General Concepts and Associated Terms (VIM), JCGM 200:2012, 3rd ed.; 2008 version with minor corrections;
BIPM: Sèvres, France, 2012.
3. ISO/IEC 17025:2017; ISO/CASCO Committee on Conformity Assessment, General Requirements for the Competence of Testing
and Calibration Laboratories. September 2017 Corrected May 2018. International Organization for Standardization: Geneva,
Switzerland, 2018.
4. Saunders, P.; Edgar, H. On the characterization and correction of the size-of-source effect in radiation thermometers. Metrologia
2009, 46, 62–74. https://doi.org/10.1088/0026-1394/46/1/008.
5. Saunders, P.; Fischer, J.; Sadli, M.; Battuello, M.; Park, C.W.; Yuan, Z.; Yoon, H.; Li, W.; van der Ham, E.; Sakuma, F.; et al.
Uncertainty Budgets for Calibration of Radiation Thermometers below the Silver Point. Int. J. Thermophys. 2008, 29, 1066–1083.
https://doi.org/10.1007/s10765-008-0385-1.
6. Pušnik, I.; Grgić, G.; Drnovšek, J. System for the determination of the size-of-source effect of radiation thermometers with the
direct reading of temperature. Meas. Sci. Technol. 2006, 17, 1330–1336. https://doi.org/10.1088/0957-0233/17/6/007.
7. Nutter, G.D. Radiation Thermometers: Design Principles and Operating Characteristics. In Theory and Practice of Radiation Ther-
mometry; Dewitt, D.P., Nutter, G.D., Eds.; Wiley Interscience: New York, NY, USA, 1988; pp. 231–337.
8. Cywiak, D.; Cárdenas-García, D.; Rodriguez-Arteaga, H. Influence of Size of Source Effect on Accuracy of LWIR Radiation
Thermometers. Metrol. Meas. Syst. 2016, 23, 661–667. https://doi.org/10.1515/mms-2016-0050.
9. Saunders, P. Calibration and use of low-temperature direct-reading radiation thermometers. Meas. Sci. Technol. 2009, 20, 025104.
https://doi.org/10.1088/0957-0233/20/2/025104.
Sensors 2022, 22, 8284 17 of 17
10. Sakuma, F.; Hattori, S. Establishing a practical temperature standard by using a narrow-band radiation thermometer with a
silicon detector. In Temperature: Its Measurement and Control in Science and Industry; Schooley, J.F., Ed.; AIP Conference Proceed-
ings: New York, NY, USA, 1982; Volume 5, pp. 421–427.
11. Saunders, P. General interpolation equations for the calibration of radiation thermometers. Metrologia 1997, 34, 201–210,
https://doi.org/10.1088/0026-1394/34/3/1.
12. Saunders, P.; White, D.R. Physical basis of interpolation equations for radiation thermometry. Metrologia 2003, 40, 195–203.
https://doi.org/10.1088/0026-1394/40/4/309.
13. Bloembergen, P.; Duan, Y.; Bosma, R.; Yuan, Z. The characterization of radiation thermometers subject to the size-of-source
effect. In Proceedings of the TEMPMEKO 1996, 6th International Symposium on Temperature and Thermal Measurements in
Industry and Science, Torino, Italy, 10–12 September 1996.
14. Jimeno-Largo, P.; Yamada, Y.; Bloembergen, P.; Villamañan, M.A.; Machin, G. Numerical analysis of the temperature drop
across the cavity wall of high temperature fixed points for radiation thermometry. In Proceedings of the TEMPMEKO 2004, 9th
International Symposium on Temperature and Thermal Measurements in Industry and Science, Zagreb, Croatia, 21–25 June
2004.
15. De Lucas, J. A Simple Geometrical Model for Calculation of the Effective Emissivity in Blackbody Cylindrical Cavities. Int. J
Thermophys. 2015, 36, 267–282. https://doi.org/10.1007/s10765-014-1757-3.
16. BIPM; IEC; IFCC; ISO; IUPAC; IUPAP; OIML; BIPM Joint Committee for Guides in Metrology. Evaluation of Measurement Data
Guide to the Expression of Uncertainty in Measurement JCGM 100:2008 (GUM 1995 with Minor Corrections), 1st ed.; BIPM: Sèvres,
France, 2008.
17. Entidad Nacional de Acreditación (ENAC). Accreditation No.16/LC10.007, Rev. 17, Date 15/01/2021. Available online:
http://www.enac.es (accessed on 20 July 2022).
18. De Lucas, J.; Segovia, J. Uncertainty calculation of the effective emissivity of cylinder-conical blackbody cavities. Metrologia 2016,
53, 61–75. https://doi.org/10.1088/0026-1394/53/1/61.
19. ISO 13528:2015; ISO, Technical Committee ISO/TC 69/SC 6 Measurement Methods and Results, Statistical Methods for Use in
Proficiency Testing by Interlaboratory Comparison. Aug/2015 Corrected Oct/2016. International Organization for Standardiza-
tion: Geneva, Switzerland, 2016.
20. McMillan, J.L.; Whittam, A.; Rokosz, M.; Simpson, R.C. Towards quantitative small-scale thermal imaging. Measurement 2008,
117, 429–434. https://doi.org/10.1016/j.measurement.2017.12.023.
21. Saunders, P.; Edgar, H. Size-of-source effect correction for a thermal imaging radiation thermometer. High Temp.-High Press.
1999, 31, 283–292. https://doi.org/10.1068/htrt166.
22. Chrzanowski, K.; Fischer, J.; Matyszkiel, R. Testing and evaluation of thermal cameras for absolute temperature measurement.
Opt. Eng. 2000, 39, 2535–2544. https://doi.org/10.1117/1.1287832.