Filters Extend Thermo graphic IR Camera Usefulness

David Bursell
FLIR Systems, Inc.

Where Filters Can Help
Materials that are transparent or opaque to IR wavelengths may present problems in non-contact temperature measurement with a thermographically calibrated IR camera. With transparent materials the camera sees through them and records a temperature that is a combination of the material itself and the target behind it. In the second case, when an IR camera needs to see through a material to measure the temperature of a target, signal attenuation and ambient reflections can make accurate temperature readings difficult or impossible. In some cases, an IR filter can be placed in the camera’s optical path to overcome these problems.

Spectral Response is the Key
IR cameras inherently measure irradiance not temperature. However, a camera’s software coverts radiance measurements into temperatures by using the known emissivity of a target object and applying internal calibration data for the camera’s spectral response. The spectral response is determined primarily by the camera’s lens and detector. Figure 1 shows the spectral response of a few IR cameras with various spectral responses. The spectral performance of most cameras can be found in their user manual or technical specifications.

Figure 1. Relative response curves for a number of IR cameras.

For many objects, emissivity is a function of their radiance wavelength, and is further influenced by their temperature, the angle at which they are viewed by a camera, and other factors. An object whose emissivity varies strongly with wavelength is called a selective radiator. One that has the same emissivity for all wavelengths is called a greybody. Transparent materials, such as glass and many plastics, tend to be selective radiators. In other words, their degree of transparency varies with wavelength. There may be IR wavelengths where they are essentially opaque due to absorption. Since, according to Kirchhoff’s Law, a good absorber is also a good emitter, this opens the possibility of measuring the radiance and temperature of a selective radiator at some wavelength.

Spectral Adaptation
Inserting a spectral filter into the camera’s optical path is called spectral adaptation. The first step of this process is to analyze the spectral properties of the target you are trying to measure. For common materials the data may be available in published data. Otherwise, this requires analysis with a spectrophotometer. (The camera manufacturer or a consulting firm may supply this service.) In either case, the objective is to find the spectral location of a band of complete absorption for the target that falls within the IR camera’s response curve.

Microbolometer detectors have rather broad response curves so they are not likely to present a problem in this respect. However, adding a filter decreases overall sensitivity due to narrowing the camera’s spectral range. Sensitivity is reduced approximately by the ratio of the area under the filter’s spectral curve to the area under the camera’s spectral curve. This could be a problem for microbolometer systems, since they have relatively low sensitivity to start with and a broad spectral curve. Using a camera with, for example, a QWIP detector will provide greater sensitivity with a narrower spectral curve. Still, this narrow range may limit the application of such cameras for spectral adaptation.

Ultimately, an optical (IR) filter must be selected that blocks all wavelengths except the band where the target absorbs. This ensures that the target has high emissivity within that band.

Selective adaptation can also be applied to gases. However, a very narrow filter might be required for selecting an absorption “spike” in a gas. Even with proper filtering, temperature measurement of gases is difficult, mainly due to unknown gas density. Selective adaptation for a gas has a better chance of success if the objective is merely gas detection, since there are less stringent requirements for quantitative accuracy. In that case sensitivity would be more important, and some gases with very high absorptance might still be measurable.

Spectral adaptation could also be applied the opposite way, i.e., selection of a spectral band where the transmission through a medium is as high as possible. The purpose would be to enable measurement on a target by seeing through the medium without any interference. The medium could be ordinary atmosphere, the atmosphere of combustion gases inside a furnace, or simply a window (or other solid) through which one wants to measure.

Filter Types
The simplest filters are broadband neutral density types that are used merely to reduce optical transmission and prevent detector saturation at high temperatures. While necessary sometimes, this is not spectral adaptation.

In spectral adaptation, filters are used in order to suppress or transmit certain wavelengths. For discussion purposes, filters can be described as Short-pass (SP), Long-pass (LP), Band-pass (BP), and Narrow band-pass (NBP). See Figure 2. SP and LP filters are specified with a cut-on and a cut-off wavelength. BP and NBP filters are specified with a center wavelength and a half-width (half-power) wavelength, the latter being the width where spectral response has decreased to 50% of its maximum.

Figure 2. Response curves for different types of filters.

For temperature measurements on transparent materials, the filter selected must provide a band of essentially complete absorption. Incomplete absorption can be used, at least theoretically, provided that both absorptance and reflectance are known and stable at the absorption band. Unfortunately, absorption often varies with both temperature and thickness of the material.

An example of applying a NBP filter to the measurement of polyethylene film temperature is shown in Figure 3. The blue curve in the figure shows the absorption band of polyethylene film. The red curve shows the transmittance of a 3.45µm NBP filter, which is designed to match polyethylene film. The green curve shows the resulting transmission through film plus the filter. This curve, running just above the zero line, indicates an excellent filter adaptation, i.e. the film appears to be opaque to the camera, and no background radiation would disturb the measurement of film temperature.

Figure 3. Application of a NBP filter to achieve nearly complete absorption and high emittance from polyethylene film, allowing its temperature measurement.

Filters can also be classified according to their application temperature. Traditionally, cold filters, filters that are stabilized at or near the same temperature as the detector, are the most accurate and desired filters for thermal signatures. Warm filters, filters screwed onto the back of the optical lens outside of the detector/cooler assembly, are also commonly used but tend to provide more radiometric calibration uncertainty due to varying IR emission with ambient temperature changes.

Once a filter is selected for use with a particular camera, the camera/filter combination needs to be calibrated by the camera manufacturer. Then the performance of the system should be characterized since accuracy and sensitivity will be affected due to a reduction in energy going to the detector.

Transparent Material Measurement Techniques
Production of sheet glass and thin plastic film requires fairly tight temperature control to maximize production quality and yield. Traditionally, temperature sensors have been embedded at the orifice of the extruder, which provides rather coarse information about sheet/film temperature. An IR machine vision system can make non-contact temperature measurements and supply more usable data on the material as its extruded. However, as described above, an appropriate filter is needed for the IR camera to make the target material appear opaque.

To ensure that the proper filter was selected, spectral response curves for the camera/filter system can be created by the camera manufacturer. (See the green curve in Figure 3.) In fact, this is generally required for permanent cold filter installations to validate filter response. Otherwise, (with supportive spectral data) the user can proceed by checking emissivity. This is a verification of emissivity efficiency for the overall system response, including the target material and camera with installed filter. Recalling Kirchhoff’s law,
rl + el + tl = 1,  or  el = 1 - tl - rl,
it is clear that in order to get an emissivity value, transmittance and reflectance at the pass band of the filter must be known. The transmittance, tl, can be taken directly from a transmission diagram like the one in Figure 3 (a value of about 0.02 in that example).

Reflectance is less easy to characterize and usually is a function of material thickness . However, a transmission diagram like the one in Figure 4 provides some indication of this parameter’s value. Using the blue curve for the thinnest polyethylene material in Figure 4, which has the lowest absorption, the transmission between absorption bands is seen to be approximately 90%. If there were no absorption bands at all, we could conclude that the reflection would be 10 %. Since there are some narrow absorption bands under the curve, we can estimate the reflection to be 8 % in the spectral regions where absorption is very low. However, we are interested in the reflectance where the absorption is high (i.e., where the material appears to be opaque).

To estimate the reflectance of this polyethylene film, we must first make the reasonable assumption that its surface reflectance stays constant over the absorption bands. Now recognize that the 8% value is the result of reflections from both sides of the film, i.e., approximately 4% per surface. At the absorption band, however, since the absorption in the material is almost complete, we get reflection only on one side. Thus rl = 0.04.

Figure 4. Transmission bands for polyethylene films of three different thicknesses.

From this rl, and the tl value obtained from the transmission graph (Figure 3 in this example), emissivity can be calculated:
el = 1 – 0.02 – 0.04 = 0.94.
This value is entered into the camera’s measurement database before having it calculate the temperatures from radiance observations.

Figure 5. Transmission curves for a common industrial glass in five thicknesses from 0.23 to 5.9mm.

Sheet and plate glass production has similar temperature measurement requirements. The most common industrial varieties are variations of soda-lime-silica glass. Although they may vary in composition and color, their spectral characteristics do not change much. Looking at the spectral transmittance of such a glass for different thicknesses (Figure 5), one can conclude that IR temperature measurement must be restricted to wavelengths above 4.3µm. Depending on glass thickness, this may require either a short wavelength (SW)or long wavelength (LW) camera/detector. SW cameras cover some portion of the spectrum from 2-5μm, and LW cameras cover some portion within 8-12μm.

In selecting a filter, the temptation might be to go for a LP type with a cut-on wavelength near the point where transmittance drops to zero. However, there are other factors to consider. For example, LP filter characteristics can interfered with the negative slope of the spectral response curve of thermo-electrically cooled HgCdTe (MCT) detectors, which are used in both SW and LW cameras. A better choice may be a NBP filter.

Figure 6. Two alternative filters for glass measurement with a SW camera.

In Figure 6, transmission characteristics of a glass, a SW camera and two filters are superimposed. The green curve represents the LP filter response curve, whereas the NBP filter response is shown in blue. The latter was selected for the spectral location where glass becomes “black”, and has a center wavelength of 5.0µm.

Figure 7. Reflectance of a common glass at normal (perpendicular) incidence.

The reflectance of this glass is shown in Figure 7. Note the peak between 8 and 12µm, which has to be avoided when using a LW camera to measure the glass.

Another consideration is the camera’s viewing angle, because glass reflectance can change with angle of incidence. Fortunately reflectance does not change much up to an angle of about 45° relative to normal incidence (Figure 8).

Figure 8. Glass reflectance as a function of camera viewing angle relative to normal incidence.

From Figure 8, a value 0.025 for the glass reflectance is valid when using either the 4.7µm LP filter or the 5.0µm NBP filter (Figure 6), because they both operate in the 5µm region. Consequently a proper value for the glass emissivity in those cases would be 1 – 0.025 = 0.975.

Transmission Band Applications
There are many applications where the user needs to find a spectral band where the medium through which the camera is looking has minimum influence on the measurement. The target of interest is at the end of a measurement path on the other side of the medium. The medium is in most cases ordinary atmosphere, but it could also be a gas or a mixture of gases ( e.g., combustion gases or flames), a window, or a solid semitransparent material.

In most cases, IR camera manufacturers have anticipated the atmospheric attenuation problem. Camera manufacturers typically add a filter that reduces measurement errors due to inaccurate and/or varying atmospheric parameters by avoiding absorption bands of the constituent gases and water vapors. This is especially needed at long measurement distances and shorter wavelengths. For SW cameras, an appropriate filter utilizes the atmospheric window between the absorption bands of H2O+CO2 around 3µm, or CO2 at 4.2 µm.

Atmospheric effects on an LW camera are much less, since the atmosphere has an excellent window from 8 to 12µm. However, cameras with a broad response curve reaching into the SW spectrum may require an LP filter. This is particularly true for high temperature measurements where the radiation is shifted towards shorter wavelengths and atmospheric influence increases. An LP filter with a cut-on at 7.4µm blocks the lower part of the camera’s response curve.

An interesting transmission band application is temperature measurements on a gas-fired furnace, oven, or similar heating equipment. Objectives could be measurement of flame temperature, or measurement of internal components through the flames. In the latter case, an unfiltered IR camera will be overwhelmed by the intense radiation from the flames, making measurement of the much weaker radiation from internal objects impossible. On the other hand, any transmission through the flames from cooler internal objects will make flame temperature measurements inaccurate.

The flame absorption spectrum in Figure 9 reveals the spectral regions where these two types of measurement could be made. There is very little radiation from the flames in the 3.9 µm area, whereas there is a lot of radiation between 4.2 and 4.4 µm range. The idea is to employ filters that utilize these spectral windows for the desired measurements.

Figure 9. Flame absorption spectrum of a gas-fired furnace with two types for filters for different measurement applications.

For measurement of internal components, you need to avoid strong absorption bands because they will attenuate the radiation from the target object, and they will also emit intensely due to the high gas temperature thus blinding the camera. Although gas-fired combustion gases consist mostly of CO2 and water vapor, an atmospheric filter is unsuitable because gas concentrations and temperatures are much higher. This makes the absorption bands deeper and broader. A 3.9µm flame filter is needed for this application . See Figure 9. This is a BP filter transmitting between 3.75 and 4.02µm. With this filter installed the camera will produce an image where the flames are almost invisible and the internal structure of the furnace is presented clearly (Figure 10).

Figure 10. FLIR ThermaCAMÔ image of furnace tubes with flame filter to allow accurate temperature measurement.

To get the maximum temperature of the flames, a 4.3µm CO2 filter will show they are as high as 1400ºC. By comparison, the furnace walls as seen with the flame filter are a relatively cool 700ºC.

Conclusions
Filters can extend the application of IR cameras into areas that might otherwise restrict their use. Still, some preliminary spectrophotometer measurements may be needed on the targets and media of interest if spectral information cannot be found in IR literature. Once a filter is selected and installed, the camera/filter system should be calibrated by the camera manufacturer. Even with a well-calibrated system, is a good idea to avoid errors by avoiding spectral regions of uncertain or varying absorption relative to the camera/filter system response spectrum.