Gas Detection using a High Quantum Efficiency, 2-D JR InSb Staring Array
`
`David S. Smith and Jeffrey Moore
`Cincinnati Electronics Corporation
`7500 Innovation Way, Mason, Ohio 45040
`
`ABSTRACT
`For many terrestrial JR imaging applications, speed and sensitivity requirements can be met by
`several sensor technologies, including PtSi staring arrays, second generation scanned MCT
`arrays, InSb, and MCT staring arrays. However, for high quantum efficiency staring arrays,
`MCT and InSb in particular, much of the available signal remains unused because of well-fill
`restrictions. In narrow spectral band imaging applications the available photon flux is
`dramatically lower and high quantum efficiency becomes critical. One such application is
`detection and imaging of gases. High quantum efficiency arrays allow the use of narrow band
`filters centered on the absorption band of the gas, thus enhancing the ability of the imager to
`discriminate between background and gas emission. Results of imaging studies performed on a
`camera with a narrow band filter for gas detection are presented and analyzed.
`
`Keywords: infrared, infrared imaging, infrared detectors, gas detection, indium antimonide
`
`As infrared imaging detector development moves toward greater resolutions and wider fields of
`view, applications other than simple JR imaging are becoming possible as well as more
`intriguing. Uses of JR cameras to produce images of gases with emission (or absorption) lines
`within the passband of the camera have already found many applications. It is possible to
`produce images of exhaled air from the lungs without great effort. The question to be considered
`here is how does one produce the highest possible gas detection capability by narrowing the
`spectral region that the JR camera sees to only the passband of the gas to be detected. The
`obvious tradeoff here is that once the camera is optimized for use in only one narrow passband it
`cannot be used for other gases, unless they happen to be in the same passband. In order to bound
`the problem two approaches have been used here. Images of a gas in the narrow passband in
`which it emits have been obtained. Noise and uniformity measurements taken for the camera
`under controlled conditions indicate the maximum sensitivity for gas detection that can be
`obtained with the camera as studied here. It should be pointed out that this study is preliminary
`and in no way has optimization of the system for this purpose been performed.
`
`3
`
`4
`
`5
`W avelength (microns)
`Fig. 1 Atm ospheric transmission, -112 meter, 50% hum idity
`
`6
`
`7
`
`1 .0
`
`0.9
`
`0.8
`
`0.7
`
`0.6
`
`0.5
`
`C0 U
`
`)
`U)
`
`EU
`
`)C C
`
`U
`
`0-8194-191 1-7/95/$6.O0
`
`SPIE Vol. 2552 / 677
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00001
`
`

`

`Shown in figure 1 is a transmission curve over approximately a 1/2 meter range for normal
`laboratory air at about 50 percent humidity. (This curve was obtained in our lab using an FTER
`spectrometer set to 4 cm1 resolution.) Notice that over even short distances that the absorption
`band of carbon dioxide at 4.26 microns is rather deep and narrow. Since carbon dioxide is easily
`obtained in tanks and is relatively harmless (in properly ventilated rooms) it was decided to do
`these preliminary studies with CO2.
`
`A very narrow passband filter was obtained in the spectral region of the CO2 absorption band.
`Figure 2 is an overlay of the cooled filter passband used in the camera with the absorption band
`of carbon dioxide. The filter obtained was virtually ideal for the detection of CO2. Also notice
`in figure 2 that a significant portion of the light emitted by an object (in this case CO2 gas) will
`be absorbed by the atmosphere on its way to the camera. This point we will discuss in more
`detail later.
`
`4.0
`
`4.1
`
`4.2
`
`1 .0
`
`C0
`. 0.8
`')C
`F- 0.6
`(I)0
`E
`
`EC
`
`0.4
`
`G)
`
`U.
`
`0.2
`
`0.0
`
`4.3
`Wavelength (microns)
`Fig. 2 Atmospheric Transmission convolved with Filter Transmission
`
`4.4
`
`4.5
`
`4.6
`
`In order to produce a wide range of images, and to provide a basis for "calibration " the CO2 gas
`was passed through a heater, thereby producing images of the gas flow at room temperature and
`above. Obviously, as the gas was heated it became easier to detect. Some estimates of the
`emissivity of gas will be used to provide a comparison between theory and experimental results.
`Shown in figure 3 is a composite image of "leaking" carbon dioxide at various temperatures.
`Two temperature sensors were placed inside the gas supply tube to provide an estimate of the gas
`temperature. One sensor was placed near the gas entry point (from the heater) and one near the
`exit point of the gas from the tube. The temperature sensor near the gas exit point consistently
`read a few degrees lower that the entry sensor so it is likely that the gas as shown in the
`following pictures was somewhat cooler than the exit temperature sensor indicated. The gas
`flow rate was on the order of 10 std-ft3/hr (relative to nitrogen). The flow could barely be felt on
`the palm of the hand.
`
`678 ISPIE Vol. 2552
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00002
`
`

`

`22 deg C = Tgas
`
`39deg.C=Tga
`
`Figure 3. Composite image of carbon
`dioxide gas escaping from the end of a tube
`at 10 std-ft3/hr. Three different gas
`temperatures, 22, 39, and 56 °C are shown
`here against a background blackbody
`temperature of 40 °C.
`
`Many more images of CO2 gas at temperatures intermediate to those shown in figure 3 were
`acquired. Figure 4 is a plot of signal levels from the areas of the images just outside the end of
`the tube (where the gas is hottest and most concentrated) and from just above the gas plume
`where the background is maintained at a constant 40 °C. The signal levels measured using the
`camera at some distance away from the blackbody will be lower than they would be when close
`to the blackbody as a result of absorption in the atmospheric CO2 between the camera and the
`heat source. In fact it appears that approximately one quarter of the radiation in the camera
`passband was absorbed before
`
`--
`--
`
`-
`
`1i_
`
`-
`
`Figure 4. Digital signal level
`from gas emission and 313 K
`background regions for the images
`shown in figure 3 and others. The
`3 1 3 K (40 °C) background level near
`23,000 units corresponds to a 76 %
`atmospheric transmission. The gas
`and blackbody were approximately 1
`meter from the camera. The thin
`horizontal line is the camera
`calibration for 40 °C at 0-distance.
`The thin line close to the gas data is
`a theoretical fit as explained in the
`text.
`
`40
`
`35
`
`:::c: 30
`ccc:ci
`05225
`, ..a
`
`20
`
`E -I
`E 1
`
`4"'$i
`
`:
`—-——i—
`ic9i!ZF
`
`-'
`
`'
`15
`295 300 305 310 315 320 325 330
`Temperature (K)
`
`SPIE Vol. 2552 I 679
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00003
`
`

`

`Figure 5a. Images of CO2 gas
`escaping the end of a 5 mm dia.
`tube at 10 ft3/hr. The appoximate
`gas temperatures are shown beside
`each sub-frame. As the
`temperature decreases the gas
`becomes less visible as long as the
`display scale is kept the same.
`Figure 5b shows how the image
`can be processed to increase
`contrast. The image in 5b is the
`entire frame corresponding to the
`subframe at the bottom of figure
`5a.
`
`52 °C
`
`48 °C
`
`40 °C
`
`33 °C
`
`27 °C
`
`22 °C
`
`Figure 5b. CO2 gas near room temperature
`escaping the end of a tube at a low rate. This
`image is histogram projected to display a range
`of roughly 3-4 bits, 1 part in 10. Even though
`the gas is near room temperature it is still
`easily detected using the InSb-based camera
`with a narrow-band filter. Background is
`uniform and at room temperature.
`
`reaching the camera. Reference to figure 2 shows that this level of absorption is consistent with
`the atmospheric transmission measured in the lab. (All of the images shown here were acquired
`with the camera approximately 1 m from the gas leak.) A "fit" to the gas emission data was
`obtained by adding 34 percent of the 40 deg. C background blackbody to 66 percent of the
`relative emitted energy from the gas. The sum was scaled to match just the blackbody emission
`at 40 C. This is expressed in the following equation where Ta is the transmission of the
`atmosphere between the blackbody/gas and the imager. Tg is the transmission of the gas itself
`and Eg is emissivity of the gas. K is an arbitrary scaling constant.
`
`Signal (DU' s) = K * Ta * [Tg * BB(313 K) + Eg
`
`*
`
`BB(Tgas) 1.
`
`680 / SPIE Vol. 2552
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00004
`
`

`

`These results suggest that the emissivity of the gas is high even for such a small diameter plume.
`(No attempt was made to calculate absolute emission and then predict camera signals using
`detailed camera characteristics such as quantum efficiency and electronic transfer functions.)
`"Random" variation in the background level and the signal levels from the gas at the different
`temperatures is most likely a result of changing CO2 concentration in the atmosphere. (A person
`was in the room where the images were acquired.) This supposition is supported by the linearity
`and uniformity measurements of the camera using a calibrated wide-area blackbody to provide
`spatially uniform calibration fluxes at a range of temperatures.
`
`Low spatial nonuniformity as well as good linearity are crucial to obtaining images as good as
`those in figure 5. The measured stability and the unformity of the imager are well beyond what
`is required to produce the images in those figures. Figures 6 and 7 show the measured linearity
`
`Camera C alib ration w /C 02 filter
`
`4000
`
`3500
`
`3000
`
`2500
`
`2000
`
`1500
`
`1000
`
`500
`
`0
`
`1
`
`2
`
`3
`7
`4
`5
`6
`Calculated Relative Flux through filter (rd.)
`
`8
`
`9
`
`Figure 6. Digital signal versus calculated relative flux through
`the narrow band filter used to obtain the images presented here.
`Linearity is good to better than 0.1 percent absolute for this filter/
`integration time combination.
`
`and the noise equivalent temperature difference for the camera with the CO2 filter in place. It
`should be noted that the NETD measured for the narrow CO2 band is almost as good as that
`normally measured for this imager with a much wider bandpass filter. This was accomplished by
`integrating the signal for almost an entire frame time instead of using only a small fraction of the
`frame as is normally the case for the wide bandpass filter.
`
`SPIE Vol. 2552/681
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00005
`
`

`

`fI
`
`1I
`1.
`1/I
`•/
`I
`
`I
`I
`I
`
`I
`
`70 - ______________________________________
`
`—.--- Temperature (C) v Spatial NEDT
`_.u—-. Temperature (C) v Temp. NEDT
`
`!
`
`60 -
`
`50-
`
`40-
`
`,
`c
`3O-
`.;
`).
`0
`z
`
`20-
`
`.
`_\
`'
`\\\
`
`I5I /
`I / I
`
`N
`
`S
`
`I
`
`10 —
`0
`
`I
`10
`
`I
`
`I
`40
`20
`30
`Background Temperature (C)
`
`50
`
`60
`
`Figure 7. Spatial and temporal noise equivalent temperature
`differences as a function of background (calibration) temperature.
`Spatial uniformity is just as important as low temporal noise.
`
`In this regard, an important aspect of an JR camera' s application to gas detection that should be
`discussed is the quantum efficiency. If the detectors in the camera used to take the data in figure
`5b had only 5 percent quantum efficiency (rather than approximately 80 percent) we would not
`have been able to easily detect the presence of the gas down to room temperature simply because
`the signal-to-noise ratio would not have been high enough. Lower Q.E. cameras would not be
`able to produce these results simply because the signal in such a narrow spectral band is too
`small for cameras (e.g PtSi and uncooled imaging systems) that require essentially the full
`spectrum of radiation (and possibly full-frame integration) to produce high sensitivity images.
`
`The NETD measurement results in figure 7 are for a controlled set of conditions with the imager
`viewing a uniform blackbody. It says that if the blackbody changes temperature in one region of
`the image we will be able to detect a change of about 0.02 K. If the viewed object were a
`graybody instead of a blackbody we would expect to detect temperature changes of 0.02 / E
`
`682 ISPIE Vol. 2552
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00006
`
`

`

`where B is between 0 and 1, the emissivity of the graybody. (As B approaches 0 the object is
`becoming reflective instead of emissive.) Our results above suggest that the CO2 gas studied
`here is almost thick enough to act as a good graybody (almost a blackbody) so the minimum gas
`temperature difference measureable should be roughly the same as the NETD measurement
`provides. The slope of the gas signal curve in figure 4 is approximately 0.002 K per digital
`signal unit. The measured NETD (see figure 7) is approximately 0.02 K. Using a slope of 0.002
`K/DU this implies a noise value of 10 digital signal units out of 65535. In order to predict
`whether a given gas leak could be detected with such an imager one needs to take into account
`the size of the leak's image on the detector focal plane. The more pixels the image covers the
`better the spatial averaging, making it easier to detect its absorption/radiation effect, up to the
`limit where too much contrast within the image is lost.
`
`If we assume that the emissivity of the gas to be detected is determined by the concentration of
`the gas then the detectable temperature difference should be approximately
`
`Temperature difference
`
`NEDT I Absolute Gas Conc.
`
`By this reasoning CO2 concentrations on the order of 0. 1 % at a temperature difference of
`approximately 12 K should be detectable assuming we use emissivity alone. If we take into
`account the contrast generated by the absorption of the gas lower temperature differences should
`be detectable. Hot gases escaping from pipes or tubing are ideal candidates for this type of
`detection even if the leak itself is not in the field-of-view of the camera. In figure 8 images of
`CO2 gas mixed with N2 gas are shown escaping the end of the same tube used to produce the
`images in figures 3 and 5. From top to bottom the images show the effect of increasing carbon
`dioxide gas concentration on the intensity of the signal from the gas plume. For these pictures an
`attempt was made to maintain the total gas flow at a constant, slow, but arbitrary rate. For the
`low gas concentrations the signal increases with increasing gas concentration. As the
`concentration passes about 25-50 percent the rate of increase in signal appears to possibly level
`off indicating that an emissivity of 1 is being approached for the gas. A one percent
`concentration is easily detected using only a temperature difference of 8 K between the gas and
`the room temperature background. As pointed out above the noise level is on the order of 10
`digital units out of 65535 for full scale. The signal level for the 1 percent concentration image of
`figure 8a is 135 digital units (See figure 8b), approximately 10 times the noise level. This means
`that we could have detected gas concentrations as small as 0.08 percent at only a temperature
`difference of 8 K. This is slightly better than prediected above using the NEDT and the gas
`concentration. (With the gas flow gauges used in the setup for this experiment it was not
`possible to produce accurately measured mixtures of gases below 1 percent.)
`
`It should be noted again that the size of the gas stream coming from the tube is only about 5 mm
`diameter. If we were searching for longer pathlengths the "emissivity" of the overall beam
`would be increased resulting in a greater ability to detect lower concentrations. If several meters
`of gas pathlength are present it should be possible to detect concentrations of CO2 down to 10-
`100 ppm. As with any radiometric measurement process the detectability depends on the
`geometry of the situation so care must be taken not to claim gas detection capability at these
`levels for all configurations.
`
`SPIE Vol. 2552 / 683
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00007
`
`

`

`Figure 8a,b. Composite JR image of a mixture
`of carbon dioxide gas and nitrogen gas at 30
`degrees centigrade. The CO2 percentage is
`listed alongside the figure. An attempt was
`made to keep the total flow rate the same for
`the mixture throughout the sequence. It
`appears that some variation of the flow was
`ocurred vis-a-vis the point at which turbulent
`flow begins for the 50 percent and the 75
`percent mixtures. As can be seen the
`emissivity increases with increasing
`concentration. Intensity vs. concentration is
`plotted below showing the relationship
`between concentration and emissivity.
`
`I
`
`II
`
`I IIIlIJ
`
`I
`
`I IIIJ
`• •
`•_
`
`•
`
`1000
`
`I
`
`1111111
`10
`CO2 Gas Concentration
`
`1111111
`100
`
`•
`
`• I
`
`I
`
`1
`
`100
`
`>a
`
`.
`
`)-J
`C0)
`
`1%
`
`2%
`
`5%
`
`10%
`
`25%
`
`50%
`
`0
`7591
`
`100%
`
`While a full calibration of the camera for purposes of gas detection was not attempted here, we
`have been able to show that small temperature or gas concentrations are detectable using a high
`quantum efficiency, mid-wave imager based on InSb. Since the minimum detectable gas
`concentration or temperature change depends highly on the nature of the background
`environment, one should not draw too many conclusions from the images presented here.
`However, the fact that relative fluxes can be calculated and fit to gas emission data with sensible
`parameters is encouraging. It means that a camera calibration for narrow spectral regions can be
`meaningful, and used for quantitative measurement of gas phenomena.
`
`684 ISPIE Vol. 2552
`
`Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/15/2013 Terms of Use: http://spiedl.org/terms
`
`FLIR Systems, Inc.
`Exhibit 1009-00008
`
`