You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281492916

Calibration of infrared radiation thermometers

Technical Report · September 2013


DOI: 10.13140/RG.2.1.2623.9203

CITATION READS
1 1,089

1 author:

Hans Gustaf Liedberg


LMC Solutions (Pty) Ltd
29 PUBLICATIONS   79 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Hans Gustaf Liedberg on 05 September 2015.

The user has requested enhancement of the downloaded file.


Calibration of infrared radiation thermometers
Speaker / Author: H. Liedberg

NMISA
Private Bag X34, Lynnwood Ridge, 0040, South Africa
e-mail: hliedber@nmisa.org
Phone: +27-12-841 2753

Abstract

Infrared radiation thermometers are typically calibrated in similar ways to contact thermometers, by
comparison with calibrated reference thermometers or at thermometric fixed points. Therefore, many of
the factors affecting the uncertainty of calibration, such as calibration and drift of the reference standard
and temperature gradients in the heat source, will be familiar to all Temperature metrologists. However,
owing to the optical nature of the measurement, some other uncertainty contributions must be
considered, notably the size and emissivity of the target. This lecture will discuss the latter components
and will demonstrate practical methods of estimating their magnitude and handling calculations related
to them.

Introduction

The temperature measured by a radiation thermometer depends on the amount of radiation emitted and
reflected into it, by objects within its field of view. This clearly depends on what it “sees” (in other
words, how big is its field-of-view?), how black its target is (how much radiation is emitted or reflected
by it?), and, if the target is not very black, what other radiation sources are in the surroundings. This
paper discusses the practical consequences of these factors, and methods that a calibration laboratory
should implement in order to deal with them.

Size of source

The target size specified by a radiation thermometer manufacturer is typically 90 % of the actual field-
of-view [1]. However, this assumes that the instrument is still “as new”: if the lens is dirty or scratched,
the thermometer may generate a lot more of its signal from outside the nominal target area. So, it is
never advisable to rely solely on the manufacturer’s field-of-view specification: it should always be
evaluated how stable the reading remains, upon small changes in orientation or target size. At each
calibration temperature, sweep the thermometer a little left and right, up and down, and look for
a stable plateau in the middle: no plateau means that the source is too small [1].

At the temperature furthest from ambient, change the distance to the target by a recorded
amount, and record the change in reading. Use this value, scaled according to the difference from
ambient temperature, as a component in the uncertainty budget. (At room temperature, there is no
size-of-source effect, as the target and the surroundings are at the same temperature. The effect becomes
larger, the greater the difference between the amount of radiation emitted by the intended target and that
emitted by the surroundings.) For example, for a thermometer calibrated from 20 °C to 400 °C, at 0.6 m
from a source of 50 mm diameter, you might vary the distance at 400 °C, measuring also at 0.8 m from
the target. Say that the reading drops by 2.6 °C with this 0.2 m increase in distance: you might estimate
that the user can duplicate your viewing conditions to with 0.1 m, so that you include a component of
0.1m (T − 20°C )
2.6°C ⋅ ⋅ as the half-width of a rectangular distribution (i.e., coverage factor
0.2m (400°C − 20°C )
k = √3) at the calibration temperature T. As can be seen, this is a “rough and ready” approach: an

{1139430221325956549}.doc Page 1 of 6
appropriate balance must be struck between the time (and cost) of the calibration and the amount of
knowledge of instrument behaviour that is provided to the client.

Be careful when calibrating radiation thermometers below ambient! The amount of radiation at low
temperatures is so small that the effect of the surroundings may be overwhelming, if the target is not
large enough. For example, dropping to -50 °C (70 K below ambient), the amount of 10 µm radiation
has decreased by almost five times; to increase by the same amount from ambient, the temperature must
rise to 150 °C (130 K above ambient).

Always report both the diameter of the target and the distance to the target in the calibration
certificate. Both are needed to properly define the viewing conditions.

Long-wavelength (10 µm), low-temperature thermometers typically require much larger sources than
high-temperature (1 µm) models. For this reason, calibration of a Cyclops 152 (operating at 1 µm, with
distance to spot ratio = 170:1) using a 25 mm diameter cavity at 1400 °C will be less prone to size-of-
source error than calibration of a Fluke 63 (operating at 10 µm, with D:S = 12:1) using a 50 mm
diameter cavity at 400 °C, although both the size of the target and the difference between ambient and
target radiation levels seem better in the latter case.

Change in reading with size of source (viewing 50 mm diameter cavity at 250 °C)

5
Indicated - actual temperature (°C)

-5

-10
0 0.5 1 1.5 2 2.5
Distance to target (m )

60x60 pixel, 12.5°x12.5° imager Cyclops 33

Figure 1. Variation in the maximum temperature indicated by an entry-level thermal imager and a good-
quality spot thermometer, both operating in the wavelength range 8 µm to 14 µm, with (instantaneous)
fields of view 0.2° and 1°, respectively. (Note that the Cyclops 33 spot thermometer could only focus at
distances of 0.9 m and longer. Therefore, its first indication of insufficient target size is the 2 K drop
from 1.5 m to 2.1 m distance.)

Figure 2. Images recorded by the 60 x 60 pixel, 12.5° x 12.5° thermal imager at distances of 0.15 m,
0.3 m, 0.6 m, 0.9 m, 1.2 m, 1.5 m and 2.1 m. Note that at distances of 2.1 m, 1.5 m and 1.2 m, the
number of pixels within 1.5 K of the maximum reading was 4, 13 and 25, respectively. Despite these

{1139430221325956549}.doc Page 2 of 6
apparently “large” clusters of hot pixels, the maximum readings in these cases were still 13 K, 8 K and
6 K below that at the closest distance.

Figures 1 and 2 demonstrate the size-of-source effect for an entry-level thermal imager with
Instantaneous Field Of View (IFOV), the angle supposedly seen by a single pixel, of 0.2°, and a good-
quality (Cyclops 33) spot thermometer with a nominal 1° angle of view (equivalent to D:S ~ 60:1).
Although the thermal imager appears to have better spatial resolution, its maximum reading drops much
more with increasing distance to target (approximately -1.5 K for every increase, 0.15 m → 0.3 m →
0.6 m → 0.9 m → 1.2 m) than does the spot thermometer, and it does not exhibit a plateau region to
indicate that the target size is sufficient. Although its accuracy specification is 2 % of reading (i.e., 5 K
at 250 °C), and all relevant pixels read within 2.4 K of the maximum reading when the target was almost
filled, it is not clear to the author that this is a sufficiently large source to calibrate such an instrument:
more experience is needed, particularly of imagers with better spatial resolution, to better understand the
size-of-source effect (SSE) as exhibited by imagers. (Perhaps the average of some number of pixels
varies less than the hottest pixel: this has not yet been studied, but appears unlikely to eliminate the large
observed differences.)

Note that both the above instruments exhibit MUCH smaller size-of-source effects than many other low-
temperature radiation thermometers. Figure 1 should be seen as close to the best expected results, rather
than typical ones.

Emissivity of source

The radiation seen by the thermometer comprises that emitted by the source, ε source ⋅ L(Tsource ) , plus that
emitted by the surroundings that is reflected by the source, (1 − ε source ) ⋅ L(Tambient ) . (Here, L(T)
represents the amount of radiation emitted by an ideal black body at temperature T. Reflectivity of
source = 1 – emissivity of source: the perfect emitter reflects nothing, and the perfect reflector emits
nothing. Both emissivity and amount of radiation will, of course, depend on the wavelength at which the
thermometer operates.) This sum is typically expressed in terms of radiance temperature, the
temperature that an ideal black body would be, in order to emit the same amount of radiation as that
emitted and reflected by the real source: L(Tradiance ) = ε source ⋅ L (T source ) + (1 − ε source ) ⋅ L (Tambient ) .
How do we use this equation to convert between apparent temperatures at different emissivities? Either
we can integrate Planck’s law over the appropriate wavelength range for the above three temperatures,
adjusting the unknown temperature (Tradiance or Tsource) until balance is achieved, or, much easier, we can
adjust the thermometer’s emissivity setting to the actual emissivity of the source, and allow the
built-in compensation to do the job.

Bear in mind the direction of the correction Tsource = Tradiance + correction:


The shinier the target, and the further its temperature below ambient, the more significant are reflections
of the radiation from surrounding objects. In other words, a greater fraction of the signal is “unwanted”,
and must be removed by applying a negative correction. Or, for T below Tambient, a real source with ε < 1
will be colder than an ideal black body, to emit and reflect the same amount of radiation.
The lower the emissivity of the target, and the further its temperature above ambient, the more radiation
is “missing”, which must be compensated by applying a positive correction. Or, for T above Tambient, a
real source with ε < 1 will be hotter than an ideal black body, to emit and reflect the same amount of
radiation.

Blacker is always better, from a calibration point of view: the higher the emissivity, the smaller the
correction that needs to be applied, and/or the smaller the uncertainty in the apparent temperature,
caused by uncertainty in emissivity. For example, at an actual temperature of 600 °C (for a Long
Wavelength InfraRed thermometer, sensitive to light of 10 µm wavelength), a target of ε = 0.98 ± 0.01

{1139430221325956549}.doc Page 3 of 6
would have an apparent temperature of 592 °C ± 4 °C, but a target of ε = 0.95 ± 0.02 would have an
apparent temperature of 579 °C ± 8 °C. (This example is realistic, as the uncertainty in the emissivity is
usually larger, the further from an ideal black body is the target.)
Note also that the effect of emissivity varies with wavelength: Were the above targets to be measured
with a 1 µm thermometer, and assuming that the emissivities at these wavelengths were still ε = 0.98
± 0.01 and ε = 0.95 ± 0.02, the apparent temperatures would be 599 °C ± 0.5 °C and 597 °C ± 1 °C, i.e.,
much less affected by the deviation from ideal black body conditions. This demonstrates why it is
usually preferred to measure at the shortest wavelength that gives enough signal. It also highlights that
corrections must be applied if a 10 µm thermometer is calibrated by comparison with a 1 µm instrument,
using a radiation source which is less black than, say, ε = 0.997, but both thermometers are set to an
emissivity other than the actual value, even if both settings are the same. (At ε = 0.997, the difference
between 10 µm and 1 µm is 1 K at 600 °C, perhaps acceptably small to remain uncorrected, rather than
the 7 K or 18 K at ε = 0.98 or ε = 0.95.)

How can you determine emissivity corrections at different wavelengths and temperatures, as in the
examples above? If you have thermometers which work in the wavelength ranges of interest, with
adjustable emissivity settings, aim them at a source which is stable at the temperature of interest
(the emissivity of the source is not important), adjust the emissivity setting and record the change
in indicated temperature. It is highly recommended to construct such tables for the emissivity of your
source, for use when calibrating instruments with fixed emissivity settings.

Below are some differences between radiance temperature (the temperature at which an ideal black body
would be, to emit the amount of radiation observed) and actual temperature (of a real source, with ε < 1)
at different temperatures and wavelengths.

Table 1. Radiance temperature – actual temperature, as a function of temperature and emissivity, for a
10 µm and a 1 µm thermometer.
8µm-14µm 0.8µm-1.1µm
ε: 0.3 0.7 0.95 0.98 0.3 0.7 0.95 0.98
t (°C)
-50 55 28 6 2
0 14 6 1 0
20 0 0 0 0
100 -51 -20 -3 -1
200 -106 -40 -6 -3
400 -215 -82 -13 -5
600 -327 -128 -21 -8 -61 -19 -3 -1
800 -445 -177 -29 -11 -89 -28 -4 -2
1000 -567 -229 -37 -15 -122 -39 -6 -2
1200 -692 -282 -46 -18 -160 -51 -8 -3
1400 -201 -65 -10 -4
1600 -247 -81 -12 -5

The only situation where you need not consider the value of your source’s emissivity, or the uncertainty
in that value, is when you calibrate a UUT radiation thermometer by comparison with a reference
standard radiation thermometer, both of which operate over the same wavelength range.

The emissivity at long wavelength infrared (around 10 µm), where low-temperature thermometers
operate, may be VERY different from that at visible wavelengths (0.4 µm to 0.7 µm) [2]. White paints
often have high emissivity (~0.95) at 10 µm, and metal surfaces blackened by a thin oxide layer may
appear shiny to a long-wavelength instrument, that effectively looks right through the thin surface layer
[3]. So, take care when constructing your own (flat plate) radiation source! (Cavities easily achieve high
emissivities at all wavelengths, but must then be long with small apertures, conditions which are limiting
in many situations.) Look, for example, at the upper line for “as-received Inconel” in Figure 3: ε ≈ 0.8 at
visible wavelengths, but ε ≈ 0.3 at 10 µm.

{1139430221325956549}.doc Page 4 of 6
Figure 3. The normal spectral emissivity for a number of metals [3]

The moral regarding emissivity is: During calibration, adjust radiation thermometers’ emissivity settings
to the actual emissivity of the source, whenever possible. For fixed emissivity instruments, apply
corrections with care, ensuring that they are applied in the right direction and that they are for the same
wavelength range as the UUT. As can be seen in Table 1, incorrect handling of emissivity can cause
huge errors!
Be especially alert for instruments that do not operate in the common 8 µm to 14 µm (“10 µm”) or
0.8 µm to 1.1 µm (“1 µm”) ranges: for example, the Calex Thermosight HT, Mikron MI-GA8+, Raytek
Raynger 3i 2M and Land Cyclops 160B (working from 250 °C upwards) operate around 1.6 µm and the
Raytek Raynger 3i G5 (working from 150 °C upwards) operates around 5 µm.

Conclusion

The optical nature of radiation thermometers causes metrologists to encounter sources of error that are
unfamiliar from contact thermometry. They must apply new techniques to evaluate and reduce these
effects. After getting a basic understanding of the relevant concepts, they must, through experimentation,
develop a sound feeling for the size of effects to be expected, and must be comfortable identifying the
influencing characteristics, such as operating wavelength, as easily as an experienced contact
thermometrist estimates adequate immersion depths in a liquid bath or an air-filled furnace “by eye”, or
focuses on different error sources (with different magnitudes) for a thermocouple and a PRT.

As interlaboratory comparisons typically use near-ideal artefacts, participation in such exercises seldom
tests a metrologist’s ability to estimate uncertainties accurately or to identify faults, in lower quality
instruments. For this reason, as someone who is new to contact thermometry should test immersion
depth for every UUT when starting out, a novice radiation thermometrist should evaluate size-of-source
effect and emissivity settings for every instrument he calibrates, until he can confidently predict which
effects will have a significant effect on the uncertainty of measurement. As seen above, these effects are
often MUCH larger than may initially be thought, and, if overlooked, may render calibration results
worthless.

{1139430221325956549}.doc Page 5 of 6
As in all metrology, a basic, practical understanding of factors influencing the measurement is
irreplaceable, and calibration results must be treated with caution until the metrologist has demonstrated
that he has gained the necessary experience. It is strongly recommended that a training plan for a
radiation thermometrist include tens or hundreds of such tests, before the metrologist starts relying on
“professional judgement” to decide if a factor is significant.

References

1. http://us.flukecal.com/blog/steps-determine-if-size-your-source-too-small, accessed 2013-09-09.


2. http://en.wikipedia.org/wiki/Luminosity_function, accessed 2013-09-11.
3. Nicholas J. V., White D. R., Traceable Temperatures, 2nd ed., Chichester, John Wiley & Sons,
2001.

{1139430221325956549}.doc Page 6 of 6

View publication stats

You might also like