High speed ultraviolet digital image correlation (UV-DIC) for dynamic strains at extreme temperatures

In extreme temperature aerospace applications, such as during hot-fire rocket testing,¹ materials must withstand high-speed loads under challenging environments.² Under such conditions, the use of strain gages is difficult because gages may either burn up or break free of a specimen’s surface,³ so non-contact strain measurements become preferable.⁴ One non-invasive and non-contacting way to collect mechanical data during these tests is Digital Image Correlation (DIC).⁵ DIC measures full-field displacements from the surface of a deforming object by applying a thin speckle pattern painted on its surface.⁶ A camera then records images of the speckle pattern as the specimen deforms.⁷ Once a test is complete, the gathered images are correlated using a computer algorithm to compute full-field displacements, the derivatives of which are then often used to compute strain.⁸ 

At extreme temperatures, there are two main challenges to perform DIC: first, the surface pattern must be able to withstand the environment without breaking loose or discoloring⁹ and second, when objects are heated above about 550 °C, they emit light in the form of blackbody radiation, which can saturate the camera sensors.¹⁰ The emitted light is known to be brighter at longer wavelengths and can be mitigated using optical band-pass filters. This filtering approach was first demonstrated in 2D by Grant et al.¹¹ and extended to 3D by Chen et al.¹² who used blue-range filters to measure strains up to 1000 and 1100 °C, respectively. The blue-filter approach was later demonstrated to 1500 °C by Novak and Zok,¹³ 2000 °C by Wang et al.,¹⁴ and ultimately to 3000 °C by Pan et al.¹⁵ More recently, it has been shown that by using ultraviolet (UV) cameras, lenses, and filters, the temperature range can be extended further, as UV light has an even shorter wavelength than blue. In a study by Berke and Lambros, blue-filtered DIC saturated at temperatures as low as 900 °C while UV-DIC remained minimally saturated at 1125 °C.¹⁶ UV-DIC has since been demonstrated to work at temperatures of at least 1600 °C,^17,18 but its upper temperature limit remains unknown. The upper temperature limit of these filtering methods is not only dependent on the wavelength of light but also on the light sensitivity of the cameras at those wavelengths.¹⁷ UV-DIC has also been demonstrated in stereo using ultraviolet diffraction gratings¹⁹ and at high magnification using a custom UV zoom lens.²⁰ More recently, UV-DIC was used to measure the coefficient of the thermal expansion (CTE) value of Inconel 718.²¹ While also using UV cameras, Zhang et al. captured full-field deformation and temperature data of C/SiC material heated up to 1500 °C.²² In all of these UV-DIC studies, the fastest frame rate of any of the UV cameras used was 43 fps, and the experiments were, therefore, limited to quasi-static measurements.

At high speeds, DIC has successfully detected full field deformation with strain rates at or above 1000 s⁻¹.²³ At these strain rates, Sutton et al. performed a room temperature tensile Kolsky bar study. Using high speed imaging, they used DIC to measure the strain along the length of the tensile specimen.²⁴ Gilat et al. used DIC analysis to prove that the deformation during a Kolsky bar test was not uniform as is widely assumed but that it varied along the specimen.²⁵ Foster et al. used high speed cameras during their experiment to observe the strain rate dependence of polycarbonate and polymethyl–methacrylate. Foster et al. used the data collected during their study to further the development of a material model for polymers.²⁶ Later, Song et al. used a similar setup to Gilat et al. to test the stress response of 4330-V steel. Song was able to use DIC and high speed cameras to determine the displacement across the entire specimen, including both the gage and non-gage regions.²⁷ Using a stereo DIC setup, Molina-Viedma et al. analyzed the modal characteristics of a multicomponent automotive system.²⁸ The light system investigated was composed of multiple materials complicating the data collected during the vibration test. In another vibration experiment, Ha et al. measured the modal response of an artificial wing using DIC.²⁹ Later, Yu and Pan used a single high speed camera with a set of mirrors to capture stereo vibration measurements.³⁰ Through the aforementioned vibration experiments, it is evident that high speed DIC introduces a distinct advantage to these experiments.³¹ 

While DIC has been used separately in many high speed or high temperature applications, there are fewer examples that involve both high temperature and high speed combined. In one study, Berke et al. utilized phase locking and DIC to measure vibrational mode shapes of rectangular plates of Hastelloy-X at room temperatures up to 600 °C.³² Using a similar setup, Lopez-Alba et al. performed a study on mode shifting in an asymmetrically heated plate to 486.8 °C (760 K). This study demonstrated that when the heating profile was asymmetric, mode shifting and jumping were present, which were not present in a uniform heating profile.³³ In a separate study performed by Santos Silva et al., a non-uniformly heated plate was studied further by changing the direction in which heating was applied. The test compared the first eleven modes at room temperature, transversely heated, and longitudinal heated to ∼550 °C.³⁴ Both vibration experiments used pulsed-laser DIC to measure the mode shapes of the plates in question. Another DIC study performed by Abotula et al. used high speed cameras to measure the dynamic response of Hastelloy X at high temperatures. The Hastelloy X specimens were loaded using a compressive shock tube apparatus at both room temperature and 900 °C. They concluded that the specimens absorbed 110% more deformation energy at 900 °C than at room temperature.³⁵ This study showed a drop-in yield strength at the higher strain rates. Song et al. used their tensile Kolsky bar to characterize iridium at high temperatures.³⁶ Using high speed DIC, they evaluated deformation uniformity under dynamic loading. They tested DOP-26 iridium alloy at 750 and 1030 °C, at strain rates of ∼1000 and 3000 s⁻¹.³⁷ 

To date, the extreme temperature capabilities of UV-DIC have yet to be demonstrated for high-speed applications. In this work, a UV amplifier is coupled with a high-speed camera to perform UV-DIC applied over a series of three experiments. Test A is a quasi-static high temperature test performed on graphite rods, demonstrating the cameras’ ability to record meaningful UV images over a broad range of temperatures. Test B is a dynamic, high-speed test performed on 304 stainless steel, demonstrating the ability to detect meaningful deformation using high speed UV images. Test C is similar to test B but includes induction heating to repeat the measurement at high temperatures. The experiments show that the UV amplifier enables high temperature and high-speed UV-DIC measurements to temperatures of at least 1600 °C and frame rates of at least 5000 fps, but the resulting measurements are significantly noisier than those made without the UV amplifier at lower temperatures.

The ability of the UV amplifier to enable UV-DIC at both high temperature and high speed is demonstrated over a series of three tests as summarized in Table I.

In test A, a graphite specimen is heated to 1600 °C in vacuum under quasi-static conditions to confirm that the UV amplifier can obtain the meaningful coefficient of thermal expansion (CTE) measurements. In test B, a 304 stainless steel plate is vibrated at its fifth resonant mode to confirm that the UV amplifier can obtain dynamic displacement measurements, which are comparable to those already established at high speed without an amplifier.³⁸ Finally, in test C, the vibration experiment is paired with an induction heater to demonstrate UV-DIC under combined high temperature and high-speed conditions.

Each test in Table I was conducted two times: first without any UV equipment, and second with a UV amplifier, UV optical bandpass filter, and UV light source. The camera used in all of the above tests was a Fastcam SA3 model from Photron (Tokyo, Japan). The camera can record at its full resolution of 1024 × 1024 pixels up to a maximum rate of 1000 frames per second or at a reduced resolution up to 100 000 frames per second. It should be noted that, because the specimen used in test A was significantly smaller than in tests B and C, it required a different lens and light source in order to achieve a meaningful resolution, as well as different cameras and amplifier settings in order to achieve sufficient brightness. The exact settings and resolutions used in each test are listed in Table II.

During the UV amplifier portions of the experiment, the high-speed camera was attached to a UVi 1850-10 amplifier from Invisible Vision Ltd. (Norwich, United Kingdom). This amplifier contains a proprietary phosphor that absorbs UV light and emits visible light, thereby effectively converting the UV light into a wavelength that the camera can see. The amplifier requires the use of two lenses: a Nikon 50 mm focal length interface lens in between the camera and amplifier, and an objective lens that is mounted outside the amplifier and is aimed at the specimen. In these tests, the objective lens was either a Tonika 100 mm AT-X Pro Macro 100 F2.8 D lens for test A or Nikon RayFact PF10545MF-UV 105 mm focal length lens (Tokyo, Japan) for tests B and C. In the visible light experiments for which no UV amplifier is used, the objective lens is attached directly onto the high-speed camera. In the UV experiments, the objective lens is additionally equipped with an XNite 330C UV bandpass filter from LDP LLC (Carlstadt, NJ, USA) to screen out visible light and ensure that only UV light enters the amplifier.

A summary of the camera and lens settings for tests A–C is included in Table II. Although the exact settings varied between tests, the settings were selected to ensure that all tests had a comparable image contrast as determined at room temperature. Specifically, the exposure time was selected to limit motion blur based on the needs of the test (relatively unimportant for quasi-static measurements in test A, but dependent on the resonant frequency in tests B and C). The camera sensitivity (which depends on the exposure time of the camera, the aperture of the lens, and the gain on the UV amplifier³⁹) was further adjusted until the difference between a typical dark pixel and a typical light pixel was about 50 greyscale values, as recommended by Reu⁴⁰ and formalized for high temperature by Thai et al.¹⁷ Representative speckle images for each of the three tests are included throughout the following subsections.

A series of high temperature tests were conducted on graphite specimens comparable to those used by Thai et al.¹⁷ The graphite specimens provided a naturally dark background on which white speckles were applied using Pyropaint 634-AL paint (Aremco, Clarkstown, NY, USA) and then cured according to the manufacturer instructions. In a vacuum, the graphite remains naturally dark up to its melting point on the order of 3000 °C, while the white paint is rated to 1800 °C. Figure 1 shows a picture of the specimen (a) and (b) and its speckle pattern (c).

Heating and loading were applied using a Gleeble 1500D thermo-mechanical system that consists of a load frame and resistive heater inside of an environmental chamber. The Gleeble environmental chamber has a window of sufficient size and material to allow either UV or visible-light images to be taken during a test. The window is supported on O-rings to maintain a vacuum seal. The vacuum ensures that the graphite specimen does not oxidize at temperatures above 400 °C.

The two camera setups used for DIC are shown mounted over the viewing window of the Gleeble system in Fig. 2. Part (a) of the figure shows the unfiltered setup, which consists of the high-speed camera and the 100 mm objective lens, but without any of the UV equipment. Part (b) shows the same camera additionally equipped with the 50 mm interface lens, UV amplifier, and UV filter. Both assemblies were supported using an aluminum T-slot frame and an optical breadboard platform from Thor Labs (Newton, NJ, USA). The unfiltered system was illuminated by a pair of blue LED ring lights from CCS, inc. (Boston, MA, USA), and the UV system was illuminated by a SOLIS-365C high power LED light from Thor Labs.

Test A followed the general procedure of Thai et al.,¹⁷ aside from the cameras and optics which were used. The specimen is heated twice: one with one K-type thermocouple in the middle and another at one end in order to determine the temperature profile, and then again with the middle thermocouple removed as to not obstruct the camera view for DIC. This has the added benefit that, although K-type thermocouples are limited to temperatures below 1250 °C, the Gleeble can receive feedback control from a thermocouple at the relatively cooler end while the center of the specimen is heated up to 1600 °C. More details on the 2-thermocouple method are given in Ref. 17. Pictures were taken throughout the heating process at increments of 100 °C up to the maximum temperature of 1600 °C. After the images were taken in both setups, the DIC data were used to compute the coefficient of thermal expansion (CTE).

In test B, the high-speed camera was used to measure the vertical displacement along the free edge of a vibrating plate specimen at room temperature. The specimen was a rectangular cantilevered plate, machined from 304 stainless steel, and mounted to the top of a Data Physics Signal Force electrodynamic shaker (San Jose, CA, USA) using a rigid steel clamping block as shown in Fig. 3. Mounted on the steel clamping block is a Dytran 3056D5T accelerometer (Chatsworth, CA, USA) to monitor acceleration and velocity at the clamp. Above the plate, a Polytec OFV-353/3001 laser Doppler vibrometer was mounted, which provided non-contacting velocity and displacement measurements. The laser vibrometer was aimed near the free edge of the plate. The accelerometer, vibrometer, and electrodynamic shaker data and control were performed by an Abacus 901 and the Data Physics software. The plate had length L = 11 in. (279.4 mm) perpendicular to the clamp, width W = 8 in. (203.2 mm) parallel to the clamp, and thickness t = 0.120 in. (3.05 mm). The dimensions were selected using equations from Furman et al.³⁸ such that the plate’s fifth resonant mode (shown in Fig. 4) had a resonant frequency of ∼400 Hz. Such plates are especially useful in vibration-based fatigue testing.⁴¹ 

The front edge of the plate was painted with a base white coat of VHT high temperature paint and a speckled with black VHT high temperature paint. Both paints are rated to 1093 °C. The VHT paint was cured per the manufacturer instructions to prevent discoloration during the high temperature portions of the experiment (test C). The specimens were then monitored using test setups shown in Fig. 3. Part (a) of the figure shows the unfiltered test setup comparable to the one used in Fig. 2(a); part (b) shows the UV light setup comparable to Fig. 2(b); and part (c) shows a representative image of the speckle pattern along the free edge of the plate. It should be noted that because the plates are so thin, in order to zoom in closely and increase the number of pixels through the thickness, the far-left edge of part (c) is actually at the center of the plate while the far-right edge of part (c) is near the node line illustrated by the solid black lines in Fig. 4, which shows a finite element contour of the plate’s fifth resonant mode. Included in Fig. 4 is a dashed box that illustrates the field of view monitored by the high-speed camera.

Next, a resonant search and dwell (RSD) test were performed at the resonant frequency of interest. During the test, the shaker kept a constant phase angle between the vibrometer and accelerometer signals to ensure that resonance was maintained. During the room temperature test, the shaker excited the plate such that the vibrometer measured an equivalent acceleration of 150 g. Throughout the test, the high-speed camera recorded images at a rate of 5000 fps. Once the test was completed, the velocity data and subsequently the displacement data derived from the laser vibrometer were exported and compared to the DIC data.

In test C, the combined high temperature and high-speed capabilities of the system are demonstrated by repeating the vibration method from test B with the addition of induction heating. The specimen was heated using an SI-10KWHF induction heater from Superior Induction (Pasadena, CA, USA) paired with a flat, pancake-type coil.³² During the high temperature vibration test, the specimen temperature was monitored using an A6751SC thermal camera from FLIR (Woburn, MA, USA). Because the induction coil and thermal camera are non-contacting, the instruments applied and monitored temperatures without affecting the resonant displacement of the specimen. Figure 5 shows an image of the equipment setup for test C with the induction heater and thermal camera labeled.

The specimen was heated to a maximum temperature of 850 °C, which occurred near the center of the free edge opposite from the clamp, as shown by the temperature distribution in Fig. 6. The resonant frequency of the plate changed significantly at high temperatures compared to at room temperature. This change in frequency is explained both by the change in stiffness of the plate at high temperature and by residual strains present due to constrained, non-uniform thermal expansion. At this temperature distribution and for the plate used in the experiment, the frequency changed from 427 Hz at room temperature in test B to 371 Hz at high temperature in test C.

Unlike the test B configuration, the laser vibrometer was not placed at the free edge of the plate. This is due to the heat of the plate burning the retro-reflective taped used by the vibrometer. Instead, the vibrometer was aimed at a known location further away from the area of heating. The displacement of the free edge was then assessed by reproducing test C in a finite element simulation using ANSYS. The plate is modeled by an 81 × 59 × 1 element mesh with quadratic elements. The model consists of three sequential analyses: a steady-state thermal analysis, a static-structural analysis, and a modal analysis. The steady-state thermal analysis is used to extrapolate the temperature distribution of the entire plate from the temperature distribution measured by the thermal camera at the free edge as shown in Fig. 6. In this analysis, shown in Fig. 7(a), the clamped edge of the plate is assumed to act as a heatsink to room temperature with free convection occurring on the upper and lower faces of the plate. The thermal distribution estimated using this analysis is then passed to a static-structural analysis to compute (1) the variation in Young’s modulus over the plate due to the temperature variation and (2) the pre-stresses in the plate, shown in Fig. 7(b), resulting from the thermal gradients. Finally, the pre-stressed static result is sent to a modal analysis to compute the high-temperature mode shape of the plate. The result of the modal analysis is shown in Fig. 7(c).

Displacements and strains were calculated by processing the high-speed images in VIC-2D, a popular DIC algorithm by Correlated Solutions (Irmo, SC, USA), which is widely used in the experimental mechanics community. Due to the differences in the specimen size, lighting conditions, and temperature range, the correlation settings between test A and tests B and C differed slightly between tests, as summarized in Table III. The test A results were used to calculate thermal strains, which were then used to compute the CTE similarly to how they have been previously demonstrated for slow-speed UV-DIC.¹⁶ The results for tests B and C were used to compute vertical displacements, which were then compared to the known mode shape.⁴² 

Figure 8 shows images collected at select temperatures for each test setup of test A, accompanied by histograms of how frequently each pixel value occurs throughout the region of interest. The unfiltered images begin to show emitted light as early as 800 °C, as indicated by the rightward shift of the histogram from a minimum value of 0 (black) toward 255 (white). The images are completely saturated by 1600 °C, as indicated by a large spike in the histogram at 255. This is consistent with other high temperature UV-DIC experiments.^16,17,43 The images collected with all UV equipment do not show any significant emitted light across the temperature range of the test, and the resulting histograms remain relatively constant for all temperatures.

A comparison between filtered and un-filtered displacement data at 800 °C, which is the highest temperature for which the un-filtered images remain minimally saturated, is shown in Fig. 9. The contours are the axial displacement along the length of the specimen. Both contours have similar magnitudes and are plotted using the same color scale, but the UV amplifier method shows slightly greater displacement than the unfiltered method and the contour is visibly noisier. An additional contour plot resulting from the image at the maximum temperature of 1600 °C is shown in Fig. 9(c). At this temperature, the unfiltered images are completely saturated, and thus, only the contour resulting from the UV-filtered images is shown. As expected, the displacements are significantly larger compared to 800 °C.

The coefficient of thermal expansion (CTE) of the specimen was then calculated using thermo-elastic Hooke’s law in three dimensions,⁴⁴ 

where ɛ_xx and ɛ_yy are the normal strain in the x and y directions, respectively, which are both measured directly by DIC; σ_xx, σ_yy, and σ_zz are the normal stresses in the x, y, and z directions, respectively; E is Young’s modulus (not needed); ν is Poisson’s ratio (assumed to be 0.20⁴⁵); CTE is the coefficient of thermal expansion; and $Δ$T is the change in temperature observed in the deformed image relative to the reference image. For unconstrained thermal expansion, ɛ_xx and ɛ_yy are expected to be equal, but the strains measured by DIC were unequal, implying some form of axial constraint (e.g., friction in the grips). Therefore, if one assumes non-zero σ_xx to allow for axial constraint but σ_yy = σ_zz = 0 due to free surfaces, these expressions reduce to Eq. (3), which is independent of stress,

Due to noise in each strain contour, the CTE measurement is also non-uniform. The mean and standard deviation of CTE were calculated at each temperature and plotted in Fig. 10. Part (a) shows the mean CTE values, while part (b) shows the standard deviations. In the case of the UV amplifier tests, some of the lower temperature measurements had a mean value that was smaller than the standard deviation, indicating that although the mean values were positive, it could not be said with statistical certainty whether the specimen was expanding or contracting. Accordingly, points with mean values less than the standard deviation are drawn using open squares, while points with means greater than the standard deviation are drawn using filled squares. The filled points are comparable to other CTE measurements in the literature of this graphite as a function of temperature.^17,45

Selected contours resulting from the room temperature vibratory test are shown in Fig. 11. The contours show the peak displacements over the course of one vibratory period, as plotted relative to a reference image at zero displacement. Part (a) is computed using unfiltered visible light, and part (b) is computed using the UV test setup. Similar to Fig. 9, both contours are similar in magnitude and are plotted on the same scale, but the contour in part (b) is visibly noisier.

Figure 12 displays, for ten select images spanning one period of oscillation, the v displacement as a function of the x location for each setup, where each line is obtained from a separate deformed image. To get a single line for each image, the v-displacement was averaged in the y-pixel direction. The thin, dashed line at the top of each plot in Fig. 12 is the average value of displacement as measured by the vibrometer for each test. This value was calculated from the Data Physics software as the magnitude of the integrated velocity. Both plots are comparable in scale, but the measurements obtained using the UV amplifier are significantly noisier.

A correlation resulting from test C is shown in Fig. 13. This correlation is a contour at the maximum displacement as recorded when the maximum temperature was 850 °C in the middle of the free edge.

Similar to the data processing of test B, the data collected during test C were used to create displacement curves as shown in Fig. 14. When compared to the previous displacement curves in Fig. 12, Fig. 14 curves show even more noise within each image.

The two sets of images shown in Fig. 8 demonstrate the need for the UV amplifier and UV filter at high temperatures. The images in the first row are obtained using the unfiltered high-speed camera, and this image set saturates at a relatively low temperature between 800 and 900 °C. In contrast, the last row of Fig. 8, which was obtained using a UV amplifier and filter, does not saturate, even at temperatures as high as 1600 °C. This suggests that the UV amplifier can potentially continue to capture data at even higher temperatures given a specimen and heating method, which exceeds 1600 °C.

The next figure shows the contours from DIC found in Fig. 9. Looking at the two contours, the shapes between them look similar, but the data using the UV amplifier are significantly noisier. This shows that both methods yield comparable results, although at temperatures low enough that the un-filtered method is still sufficient, the UV amplifier method is less precise. In Fig. 9(c), a contour collected from the UV amplifier with the specimen at 1600 °C is shown. This shows that at higher temperatures, the UV amplifier is still able to record reasonable displacements well above which the un-filtered camera saturates.

In principle, these abilities could have been shown by instead using reference and deformed image pairs taken at each temperature and then performing a correlation separately at each temperature. This approach, which was used by Thai et al.¹⁷ and by Jarrett et al.,⁴⁶ has the advantage that it can be used to better assess the measurement uncertainty in UV-DIC as a function of temperature. However, because the relative strains between the two images are nominally zero, it cannot be used to measure thermal expansion or say anything else meaningful about the material being monitored. The CTE values depicted in the figure are comparable to those measured by Thai et al.¹⁷ 

The vibrometer data shown in Fig. 12(a) show that the value measured by the laser vibrometer agrees with the peak values measured by DIC. The vibrometer was slightly lower than the DIC data, but the difference was approximately only 0.003 mm for the unfiltered test and 0.004 mm for the UV amplifier test.

The contours shown in Fig. 11 indicate that the unfiltered test B and the UV amplifier test B have a similar shape and maximum displacement. This shows good agreement between the two methods tested at temperatures low enough for which the unfiltered method remains valid. The displacement plots shown in Fig. 12 further support this conclusion. Figure 12 shows that, while there is more noise to the correlation, the half-parabolic trend between the unfiltered test B and UV amplifier test B is consistent between the two methods.

The contour from the high temperature test is shown in Fig. 13, accompanied by the displacement plots in Fig. 14. The figures show that the UV system can successfully perform UV-DIC at high-speeds and high temperatures above which the emitted light significantly impacts the unfiltered system. However, the UV results are significantly noisier when compared to the unfiltered results at lower temperatures. This is likely at least partly due to warping of the air at high temperatures. The other high temperature experiment (test A) did not have such warping because the quasi-static thermal expansion tests were performed in a vacuum (i.e., no air to warp), but the vibration system did not allow for vacuum. Others have shown that such warping can be mitigated by averaging together multiple images taken with nominally no displacement between them,⁴⁷ but this requires precise camera triggering such that each image is always taken at the same phase angle within the period of oscillation.⁴² 

For the high temperature portion of the experiment, the vibrometer had to be moved away from the plate edge during the test. As such, it required a different method to validate the displacements measured in the high speed and high temperature experiment. To do this, a modal simulation was performed using the experimental conditions obtained during the test. The results from the simulation are shown in Fig. 15, along with measured displacements from the peak deformation during test C. Despite the noisy displacement measurements, the figure shows close agreement between the DIC data and the modal simulation of the plate. This somewhat confirms that, under high temperature conditions for which the unfiltered camera can no longer view the plate, the UV amplified camera correctly detects motion, which is otherwise expected.

In environments where the light emitted by a glowing specimen saturates the camera sensor, the UV amplifier successfully helps increase the useable temperature range of the high-speed camera. In the quasi-static portion of the experiment (test A), the UV amplifier and UV filter allowed useful data collection up to 1600 °C, filtering out all light emitted by the specimen. The high-speed portion of the experiment (tests B and C) allowed data collection at a frame rate of 5000 fps, which is more than 100 times faster than other UV cameras demonstrated in the literature,²² and a temperature of at least 850 °C. The results obtained using the UV amplifier are significantly noisier than those obtained without the amplifier, but the results generally agree at temperatures low enough for which the unamplified system can still obtain measurements. Therefore, although we recommend avoiding the amplifier at lower temperatures for which the unfiltered camera remains sufficient, the UV amplifier can still enable meaningful high-speed measurements at higher temperatures for which the unfiltered camera saturates.

This study was supported by funding received from the DOE Office of Nuclear Energy’s Nuclear Energy University Program (DOE-NEUP) under Award Nos. DE-NE0008531 and DE-NE0008799. B.D.H., B.A.F., and R.B.B. would also like to thank the support from the Air Force Office of Scientific Research under Award No. FA9550-21-1-0437. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the Department of Energy Office of Nuclear Energy or the U.S. Department of Defense.

The authors have no conflicts to disclose.

The data used in this paper is available upon reasonable request by email to the corresponding author.