18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
A Performance Comparison between Classical Schlieren and Background-Oriented Schlieren
P. Manovski1,*, J. Wehrmeyer2, K. Scott2, B. Loxton1, H. Quick1, S. Lam1, M. Giacobello1 1: Aerospace Division, Defence Science and Technology Group, Australia 2: Arnold Engineering Development Complex, Arnold Air Force Base, United States of America * Correspondent author:
[email protected] Keywords: Background-oriented schlieren, Quantitative or Calibrated schlieren
ABSTRACT A performance comparison was made between a classical calibrated schlieren system and a background-oriented schlieren (BOS) system at the Defence Science and Technology Group (DST Group) Transonic Wind Tunnel (TWT) in Melbourne, Australia. Comparisons between these two techniques have been made by others. This study aims to contribute to the findings by conducting a similar investigation in a large-scale facility using the latest high pixel resolution (29 MP), high sensitivity cameras to maximise spatial resolution and the ability to resolve small refraction angles. This study also aims to investigate the effects of the conical perspective of a typical BOS setup compared with the cylindrical view of schlieren. Two configurations were investigated, firstly a standard setup and then a high sensitivity setup. The sensitivity was determined experimentally and compared with theory. The minimum detectability was determined to be of the same order, approximately ±1 arc-second for both systems. The performance of each system was then investigated by imaging the compressible flow around a cone-cylinder model for a range of Mach numbers from 0.6 to 1.2. Qualitatively, a good agreement was obtained. By translating the BOS system in the streamwise direction, at 1.05 Mach, the effect of the BOS system conical perspective was evident, a relatively strong bow shock at the tip of the cone-cylinder model could be made to disappear from the BOS image. A quantitative comparison of the measured line-of-sight refraction angle showed a good agreement near the centre of the BOS lens but away from the centre discrepancies were evident and are likely due to the BOS conical perspective. The schlieren system range was also shown to be limited.
1. Introduction Both classical schlieren and background-oriented schlieren (BOS) techniques measure refractiveindex variations and can be used to visualise the line-of-sight quantitative density gradient information. The techniques are widely used to study compressible flow phenomena, such as shock waves, expansion fans, and variable-density wakes and boundary layers in high speed flows. Standard schlieren imaging is used in many wind tunnels, such as the Defence Science and Technology Group (DST Group) Transonic Wind Tunnel (TWT) in Melbourne, Australia and has a long development history [1]. In contrast, BOS is a relatively new technique that relies on digital imaging cameras and computer-based image correlation software [2]. BOS is used in the Arnold Engineering Development Complex (AEDC) 16T (16 foot Transonic) wind tunnel, rather than classical schlieren, because of the latter technique’s requirement of optical access that would block a large percentage of the test section walls. Wall blockage is detrimental to test section flow quality for transonic wind tunnels, whose test section walls are
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
ventilated to minimize shock reflection. The DST Group TWT is often used with a half model mounted on the side wall which blocks the optical access required for schlieren. For the DST Group, BOS presents an alternative that is worth investigating. The purpose of this study was to obtain a relative performance comparison between the BOS and a standard z-type schlieren system installed at the DST Group TWT while considering the facility’s geometric constraints. In particular, the aim was to compare the ability of the two techniques in detecting small density gradients. Comparisons between these two techniques have been made by others [3, 4]. This study aims to contribute to the findings by conducting a similar investigation in a large scale wind tunnel facility using the latest high pixel resolution (29 MP), high sensitivity cameras to maximize spatial resolution and the ability to resolve small refraction angles. This study also aims to investigate the effects of the conical perspective of a typical BOS setup compared with the cylindrical view of schlieren. Two configurations were investigated, firstly a standard setup that is typically used for test campaigns in the TWT and then a high sensitivity setup. The sensitivity and minimum detectability were experimentally determined by introducing a wedged optic that produces a very small (known) angular refraction as the schlieren object. The performance of each system was assessed qualitatively and quantitatively by imaging the compressible flow around a cone-cylinder model for a range of Mach numbers from 0.6 to 1.2. 2. Experimental equipment 2.1 DST Group Transonic Wind Tunnel The DST Group Transonic Wind Tunnel is a closed-circuit continuous flow wind-tunnel. A schematic of the circuit is shown in Fig. 1. The tunnel is equipped with a Plenum Evacuation System (PES), used to pressurize and depressurize the tunnel circuit as well as improve the flow quality in the test section by removing air through the slotted test section walls, and re-injecting it downstream of the first corner after compression. The tunnels convergent nozzle can be operated in an under-expanded mode to produce supersonic Mach numbers by evacuating the test section to allow expansion of the sonic nozzle flow. The total pressure in the circuit can be evacuated down to 40 kPa or pressurized up to 200 kPa. When the test section is empty the tunnel has a Mach number range of 0.30 to 1.20. The test section is nominally 2700 mm long, 806 mm high and 806 mm wide. The sidewalls are interchangeable and can be either solid or slotted. Each slotted wall consists of six slots per wall giving an overall porosity of 4.97%. The main model support is downstream of the test section, incorporating a pitch and roll mechanism. The pitch mechanism has a position accuracy of ±0.02 deg and the roll drive is accurate to within ±0.1 deg. Further details of the model support system are in [5]. The Mach number is controlled by adjusting the main fan speed and plenum suction via two control valves (one for gross control, the other for fine control) in the PES inlet pipe, thus controlling the rate of air removed from the plenum. The Mach number is monitored by the tunnel control system from the total pressure in the settling chamber and static pressure in the plenum, and is controlled to within ±0.002 for subsonic operations and ±0.01 for supersonic [5].
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Temperature in the test section, which can be set between 30°C and 40°C, is controlled via a heat exchanger in the settling chamber to within ±2°C. Humidity is reduced to below 2000 ppmv before each test run through a dehumidifier in the PES.
Fig. 1 DST Group TWT circuit schematic
Fig. 2 Modified z-type schlieren system layout in the DST Group TWT. 2.2 DST Group TWT schlieren system
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
A classical schlieren system for flow visualisation is based on the deflection of a light beam as it passes through a change in refractive index caused by a density gradient in the wind tunnel test section [1]. A properly calibrated system can measure the magnitude of the density gradient, normal to the knife edge from the intensity of the light reaching the sensor [1]. The DST Group schlieren system employs a modified version of the z-type Toepler layout and is shown in Fig. 2. The modifications include the use of two plane mirrors (PM1 and PM2) that fold the light path to reduce the space necessary for the standard schlieren system. The system mirrors consist of two 406 mm parabolic schlieren mirrors, with focal lengths SM1 = 2517 mm and SM2 = 2533 mm. The z-type arrangement introduces astigmatism into the system, however, a combination of careful alignment and minimizing the off-axis angles reduces this to a minimum and the resultant effect in the classical setup with a knife edge cut-off is negligible. Further details of the TWT schlieren system can be found in [6]. In this study a classical setup was used with a rectangular light source slit and a vertical knife edge allowing the visualisation of the streamwise density gradients. A micrometer was used to accurately measure the cut-off percentage or unobstructed source height, a. A white LED light source (LED Engin 24 die LZP-DOCW00) was used to provide stable high powered uniform illumination during the test period. The LED was driven by a Garadsoft RT820F-20 controller, capable of providing continuous or pulsed operation, with pulse widths from 3 µs – 200 ms. An achromatic doublet imaging lens of focal length 100 mm was used to provide a field-of-view encompassing the entire schlieren window. The camera used for the schlieren imaging was LaVision Imager LX 29M with camera link interface, and provided a 12 bit greyscale pixel depth. To calculate the sensitivity for the schlieren system, the camera pixel intensities, E, is converted into a contrast or normalised intensity as per, 𝐸 − 𝐸'() ∆𝐸 𝐶= = , (1) 𝐸'() 𝐸'() where 𝐸'() , is the background image or reference intensity. The reference intensity is the image intensity of undisturbed (zero density gradient) region of the image. The sensitivity, S, of a schlieren system, also referred to as contrast sensitivity, is related to how much change in contrast, C, occurs with respect to the angle of refraction, ε, and can be shown to be equal to [1], 𝑆=
12 13
)
= 4 , 5
(2)
where f2 is the focal length of the second schlieren mirror and a is the unobstructed source height. The response of a schlieren method can be calibrated by introducing into the test section an object of known refractive angle distribution. In this case a long focal length lens as suggested by [4], was used to calibrate the schlieren system and obtain the refractive angle from the image pixel intensity. A light ray which passes through a spherical lens with focal length, f, at particular radial distance, r, will be refracted through an angle 𝑟 (3) = 𝑡𝑎𝑛𝜀 ≈ 𝜀. 𝑓
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Thus a linear relationship between the refraction angle (image intensity) and the position, r, across the lens is expected. A line of best fit can be obtained from the data with the gradient, Sexp, of the line corresponding to the experimentally measured sensitivity of the schlieren system. The integrated line-of-sight refraction angle at particular point in the schlieren image, 𝜀DE5F( , is then calculated by determining the contrast and dividing by the measured sensitivity, (𝐸 − 𝐸'() ) 𝐶 𝜀DE5F( = = . (4) 𝑆(GH 𝑆(GH 𝐸'() As the method above relies upon the reference intensity, three different means of obtaining the reference intensity were compared. The effect of using the theoretical sensitivity (Eq. 2) and the calculated sensitivity was also investigated. 2.3 AEDC BOS system The BOS technique is based on the refraction of light rays coming from an illuminated background image as they pass through a region with a variable density gradient. The effect of the density gradient is to effectively shift the pattern on the background as seen by the camera due to the refraction of the light path. The magnitude and direction of the shift is proportional to the strength and direction of the density gradient that is traversed by the light ray. A recent review of BOS techniques and the underlying principle can be found in the literature [2]. BOS involves obtaining one reference image of the background with no flow field, in this case a wind tunnel scene before (or after) a transonic flow field is present. A second image is then obtained of the flow field of interest. These two images are then compared to determine the apparent movement of the background pattern. Digital image processing is used to perform cross-correlation of the images using algorithms developed for Particle Image Velocimetry (PIV) [7]. The cross-correlation algorithms determine the relative displacement of dots on the background between the pair of images and is related to the strength of the density gradient (rather than velocity as in PIV). The BOS system for this work is designed to fit into the TWT plenum. The camera and its pulsed lighting system are located close to the test section window. By doing this, reflections from the high-intensity pulsed lights into the camera lens are avoided. Figure 3 shows a schematic of the BOS system installed in the TWT, along with pertinent dimensions and Fig. 4 shows images of the BOS setup.
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
ZD2 ZB2
ZD1 ZB1
Zi
Fig. 3 The AEDC BOS system installed in the TWT with wedge optics positioned in the centre of the test section. The BOS background dot pattern was created by spraying black paint on a retroreflective selfadhesive sheet (3MTM Scotchlite). The paint was sprayed in a direction parallel to the background and a fan was used to blow the smaller diameter paint particles onto the background surface. In this way an average dot size in the order of 1 mm diameter was achieved. The BOS system used two (of the four) Excelites FX-4400 xenon flashlamps shown in Fig. 4b with a pulse duration of approximately 5 µs. The short pulse widths allowed ‘freezing’ of the flow field and reduced image blur. The camera used for the BOS imaging was LaVision Imager LX 29M with GigE interface and provided a 14 bit greyscale pixel depth. A 60 mm Nikon Micro Nikkor lens was attached to the camera providing an adequate field-of-view and resulting in a imaged background dot size on the order of 2-4 pixels. The high powered flashlamps provided sufficient illumination for an f-stop of 32 to be used, this allowed adequate depth-of-field so that both the schlieren object and the background dots were in sharp focus. LaVision Flowmaster software (DaVis 8.3) was used to obtain the displacement fields with a final interrogation window size of 16 x 16 pixels and a 75% overlap. Thin lens theory approximations were used to obtain the relationship between pixeldisplacement and refraction angle, as outlined in [2, 3 and 4]. Assuming small angles, the line-ofsight refraction angle, εx, measured by a BOS system is related to the pixel displacement, ∆𝑥, the distance from the schlieren source to the background, ZD, the distance from the lens to the background, ZB, and the distance from the lens to the imager, Zi , by ΔxZ B Δx , εx = = (5) Z D M Z D Zi where M = Zi/ZB is the magnification factor of the background. The next step is to apply a pixelto-length calibration to convert into units of the background’s dimensional scale, rather than the imager dimensional scale. From Eq. (5) the BOS systems sensitivity is defined,
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
ZB . (6) Z D Zi The dimensions in Fig. 3 show two distances for the background from the schlieren object (here a wedged optic), the first is 458 mm (ZD1) is for the standard setup with the background placed close to the far test section window and held steady against the window’s collar. This configuration is representative of when classic schlieren cannot be used due to blocked optical access. The distance of ZD2 = 1268 mm is for the high sensitivity configuration with the background moved back as far as possible within the plenum of the TWT. The distances from the background to the BOS lens (using a thin lens approximation) are ZB1 = 950 mm and ZB2 = 1762 mm. The distances from the lens to the imager were estimated from thin lens theory, Zi1 = 64 mm and Zi2 = 62 mm. From Eq. (6) two important parameters for the BOS system sensitivity are, 1) the distance of the schlieren object from the BOS background, and 2) the size of the imaged area relative to the sensor size (i.e. the magnification). S BOS =
For this test a like-for-like comparison was desired, thus the BOS system was mounted on a rotation stage and two translation stages inside the TWT, as shown in Fig. 4. These stages allowed the BOS system components to be moved out of the light path of the standard schlieren system so that both flow systems could be used in a sequential manner during the test campaign, while maintaining the tunnel conditions on a steady set-point and ensuring minimal variability in test conditions. (a)
(b)
(c)
(d)
Fig. 4 (a) BOS Camera positioned at the schlieren window of the TWT test section. (b) BOS camera looking from inside of the TWT test section. Also shown are the wedge optic and lens used for system sensitivity tests. (c) The BOS background located against the outside of the TWT test section window from inside the test section. (d) The BOS background on translation and rotation stages located against TWT test section schlieren window.
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
2.4 BOS vibration correction One challenge to implementing the BOS technique in a production wind-tunnel environment is tunnel vibration. Even with a rigidly mounted camera, there may be some relative displacement of the camera relative to the background. In addition, the individual elements of a consumergrade lens may vibrate inside their housing. This results in a complex distortion from image to image that cannot be removed using a simple translation and/or rotation correction. To solve this problem, AEDC personnel have written an algorithm which operates on the vector field result created from the cross-correlation processing. It relies on the displacements due to vibration being small enough that the images can still be correlated using multiple passes with the desired final interrogation window size, such as 16 x 16 pixels. In 16T, the typical displacement due to vibration is less than 8 pixels, so image correlation nearly always produces a usable result. The algorithm creates a 12-parameter fit of the vector field result. This fit contains all of the large scale (global) distortions in the image typically created by the vibration of the test cell. Rigid-body rotation and translation, as well as keystoning and pincushion/barrel distortions on the scale of the image size, are removed using the fit. When these large-scale distortions are subtracted from the vector field result, what remains are the small-scale (localised) distortions typically created by density gradients such as shock waves or supersonic jets. Because all large-scale gradients are removed in this way, real density gradients which extend over a large portion of the image are removed as well. This is usually an acceptable trade-off because the flow features of interest are typically sharp, localized gradients created by a test article. 3. Sensitivity and minimum detectability The sensitivity of an instrument is a measure of the effect its input will have on its output, that is d(output)/d(input). It is instructive to compare the experimentally-measured sensitivity of each system to an expected value, based on theory and the geometry of the system. It is also informative to determine the minimum detectability, which is the smallest signal change that can be reliably measured by each system. For the two configurations investigated (standard and high sensitivity) these quantities were determined as outlined in the following section. When light goes through an optical wedge it will result in that area of light having a constant angle of refraction and therefore constant intensity for a schlieren image and a constant displacement for a BOS image. Placing the wedge optic as a schlieren object in the test section of the TWT and rotating it, will produce a sinusoidal response in intensity (brightness) for the classical schlieren system. This is due to the knife edge measuring the refraction only in a direction normal to the knife edge (for example only the horizontal direction if the knife edge is vertical). The BOS system measures both the horizontal and vertical displacements, thus the same result can be obtained from the BOS system if the horizontal (or vertical) displacements are extracted and plotted. The minimum detectability values were experimentally determined for both the schlieren and BOS systems by rotating a wedge optic and taking multiple images at 10° increments until a full revolution of the wedge was completed. By plotting the schlieren intensity (or displacement for BOS) values against the wedge rotation angles transformed into an expected horizontal
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
refraction angle, a linear relationship is expected. The minimum detectability was defined as twice the standard deviation of the difference between the measured value and this linear line of best fit. The intensity (or displacement) of the wedge at each of the 36 data points was found by averaging both in space and time. The spatial average was calculated by averaging the intensity (or displacement) within the interior 50% area of the wedge optic. This was done over a series of images to obtain an ensemble average. To find the refraction angle produced by the wedge in the horizontal direction, as the knife edge was vertical, the angle of deflection, 𝜃1 and the phase shift 𝜃TUD)V of the wedge had to be calculated. The angle of deflection represents the constant angle produced by the refraction of the light through the optic. It is related to wedge angle, 𝜃W , by 𝜃W = tanZ[
\]^ _` aZbc\ _`
,
(7)
where 𝑛, is the refractive index of the wedge. The derivation above comes from Snell’s Law and the equation can be numerically solved to obtain 𝜃1 . The wedges used for the experiments were UV grade fused silica, having a refractive index of 1.514. The horizontal angle of refraction produced by the wedge was then calculated by 𝜀G = 𝐴 sin(𝜃 − 𝜃TUD)V ),
(8)
where the amplitude, A, is the contrast (or BOS pixel displacement) for the maximum expected refraction angle (i.e. the deflection angle, 𝜃1 ) and the phase shift, 𝜃TUD)V was found by looking at the intensity response to the wedge rotation and finding the angle is required to shift the data back to a standard sine curve. To calculate the sensitivity for the classical schlieren system, the reference intensity was calculated by plotting the line of best fit of the intensity against the horizontal angle of refraction then finding the image intensity for zero angle of refraction (𝜀G = 0). This value was then used to produce the contrast, C. The gradient of the line of best fit of the contrast (or displacement for BOS) values plotted against the horizontal refraction angle corresponds to the contrast sensitivity for the schlieren system (and the sensitivity for the BOS system). 3.1 Standard sensitivity configuration The TWT standard schlieren system was setup with a 3.03 mm source slit and a cut-off of approximately 50%, resulting in a measured unobstructed source height, a = 1.52 mm. For the BOS standard setup the background was placed up against the window collar as indicated in Fig. 3. For the standard sensitivity configuration, a single wedge optic (Thorlabs model SI500P) of 10.8 ±0.5 arc-seconds wedge angle was used to provide a range of horizontal light ray deflections. The deviation angle was calculated to be 5.55 arc-seconds. Both systems were used to image the wedge optic which was mounted on a precision optics holder in the test section. The contrast and displacements are plotted against horizontal refraction angle as shown in Fig 5. Both were found to have the sinusoidal response as expected. Figure 6 shows the schlieren and BOS, contrast and displacement, against the wedge optic rotation angle transformed into a refraction angle (±5.55 arc-seconds) and the linear line of best-fit, whose slope is the sensitivity for each system. The minimum detectability defined was found to be ±0.75 arc-seconds for the schlieren system and for the BOS system it was ±1.1 arc-seconds. The experimentally-
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
determined static sensitivities were 0.01 contrast/arc-second and 0.025 pixels/arc-second, for the schlieren and BOS systems, respectively. These values are approximately 25% higher for the schlieren system and 7% lower for the BOS system, compared with their respective theoretical sensitivities of 0.0081 contrast/arc-second and 0.027 pixels/arc-second. Both experimental and theoretical sensitivities are plotted for each system in Fig. 7. Post-test the procedure was repeated for the BOS system using a pair of wedge optics (described below) with a total wedge angle of 5.2 ± 0.7 arc-seconds, this time the resulting minimum detectability was ±0.42 arc-seconds and the calculated sensitivity was 23% higher than the theoretical value. The results are summarised in Table 1. Discussion about the discrepancies is presented in the next section. 3.2 High sensitivity configuration For the high sensitivity schlieren configuration, the source height was reduced by a factor of almost three to 1.12 mm and the unobstructed source height was measured to be 0.56 mm (i.e. a 50% cut-off). According to Eq.2 this should result in approximately 2.75 times improvement in sensitivity. For the high sensitivity BOS setup the background was moved back an additional distance of 810 mm and due to refocusing of the lens, Zi2 = 62 mm and ZB2 = 494 mm (using the thin lens approximation). According to Eq.(6) this results in approximately a 1.8 times greater sensitivity, however a contrary effect of a greater background distance is the reduction in spatial resolution which was not taken into account. The same experiment was repeated but this time a pair of wedge optics (Newport Optics model 20QS20) were used and mounted in individual rotation mounts, allowing independent rotation of either wedge. One wedge optic has a 2.9 ±0.5 arc-seconds wedge angle, and the other has a 2.3 ±0.5 arc-second wedge angle. Thus the combined total wedge angle is 5.2 ± 0.7 arc-seconds resulting in a deflection angle of 2.67 arcseconds. When configured for use in the TWT, the minimum detectability was found to be ±1.26 arc-seconds for the schlieren system and ±0.69 arc-seconds for the BOS system. The experimentally determined sensitivities were 0.025 contrast/arc-seconds and 0.053 pixels/arcsecond, for the schlieren and BOS systems, respectively. These values were approximately 13% higher for the schlieren imaging and 18% higher for the BOS system, compared with their respective theoretical sensitivities of 0.022 contrast/arc-second and 0.045 pixels/arc-second. The results of these tests are summarised in Table 1. The plots of these tests are omitted for brevity. The wedge optic tests for both systems show some variability in the obtained minimum detectability as well as discrepancies between the theoretical and measured sensitivity values. In practice a range of factors contribute to the noise in the data. As the optics were rotated in the holders for each orientation there is an error in setting the angle. For BOS, the measured displacements are so small (on the order of 0.1 pixels or less) that the slightest angular change in either wedge can result in a measurement error. Likewise, with schlieren, the variations in the schlieren camera intensity counts are of the order of 10-20 counts and any slight angular change results in measurement error. The method for accurately positioning and rotating the wedge optic could be improved by incorporating motorised rotation mounts. Table 1 Summary of the theoretical and experimental determined sensitivities and minimum detectability for both schlieren and BOS systems. schlieren configuration
source height
a (mm)
sensitivity (contrast/arc-second)
Δ
minimum detectability
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
standard sensitivity high sensitivity BOS configuration standard sensitivity standard (repeat) high sensitivity
(mm) 3.03 1.12
1.52 0.56
distance to background ZD1 = 458 mm ZD1 = 458 mm ZD2 = 1268 mm
theoretical 0.008 0.022
experimental 0.010 0.025
sensitivity (pixels/arc-second) theoretical experimental 0.027 0.025 0.027 0.035 0.045 0.053
25% 13%
(arc-second) ±0.75 ±1.26
Δ -7% 23% 18%
minimum detectability (arc-second) ±1.10 ±0.42 ±0.69
The wedge angles of the optics were accurately measured using a calibrated interferometer by Zygo Corporation’s Metrology Services Division. The uncertainty of the wedge angles is relatively high and more accurate optics could be used to characterise the system. All of the dimensions used in the calculations were measured to within ±5%, so it is doubtful they would combine together to form the 18-25% difference. At this time, the differences between experimental and theoretical static sensitivities cannot be explained and will be subject of further investigation. Despite the above, the data can still be used to conservatively estimate a minimum detectability of the order of ±1 arc-seconds for both the schlieren and BOS systems as configured in the TWT wind tunnel. The theoretical minimum detectability for a schlieren system is dependent on the reference (or background) image intensity, the sensors spatial resolution and the sensors ability or resolution to measured light (pixel depth). For the tested configurations the reference intensity was approximately 1700 counts for the standard configuration and around 420 counts for the high sensitivity configuration. To obtain the theoretical minimum detectability, we consider a typical ±10 counts intensity variation due to camera noise and determine the corresponding contrast from Eq.(1) and then from Eq.(2) divide by the sensitivity. The theoretical minimum detectability for the schlieren system is 0.7 arc-seconds for the standard configuration and 1.1 arc-seconds for the high sensitivity configuration. These values are very similar to those obtained in this study. The minimum detectability did not improve for the high sensitivity configuration due to the reduced light intensity as a result of the smaller unobstructed source height. For the BOS system it is well reported that the minimum resolvable pixel displacement from the PIV cross-correlation algorithms is of the order of 0.05-0.1 pixels [7]. If we multiply the inverse of the theoretical sensitivities by 0.05 (pixels), we obtain the theoretical BOS minimum detectability of 1.85 arc-seconds for the standard configuration and 1.28 arc-seconds for the high sensitivity configuration. These values are slightly higher than what we obtained in the experimentally measured values, which suggests that our BOS system is performing well. The performance improvement may be attributed to a reduction in the displacement uncertainty after performing a spatial average of the displacement over the wedge and then averaging over a number of images.
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Fig. 5 For the standard sensitivity configuration, left: schlieren (contrast) and right: BOS signal (dispalcement), produced by rotating a 10.8 arc-second wedge optic. The expected response is a sine wave, shown by the solid curve.
Fig. 6 For the standard sensitivity configuration, left: schlieren (contrast) and right: BOS signal (displacement) response to a change in refraction angle caused by the wedge optic rotation. The horizontal axis is scaled to the assumed ±5.55 arc-seconds maximum deflection angle. The line of best fit (sensitivity) and the minimum detectability are shown.
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Fig. 7 For the standard sensitivity configuration, left: schlieren (contrast) and right: BOS signal (dispalcement), response to a change in refraction angle. The line of best fit provides the experimentally derived sensitivity and also shown is the theoretical sensitivity line. 3.3 Dynamic range comparison Both flow visualisation systems can handle much larger ranges of input signals, and a spherical lens can be used to demonstrate this range. The measurement range for a BOS system is influenced by the correlation algorithm’s search area, which is usually a significant fraction of the interrogation region, here 16 pixels. If a 25% maximum displacement is assumed, the BOS range can be on the order of ±4 pixels. With a minimum detectability, in terms of signal, of ±0.05-0.1 pixels [7], the dynamic range of a BOS system can be on the order of 160. Even greater dynamic range can be obtained using correlation techniques with adaptive interrogation techniques. For standard schlieren, the dynamic range is related to the source size and cut-off ratio. By reducing the source size, the dynamic range is lessened at the expense of increasing the sensitivity. This is demonstrated in Fig. 8 where BOS and standard schlieren images of a 10 meter focal length spherical lens are shown. The lens is CVI BK7 Spherical Plano-Convex Lens (model number PLCX-50.8-5151.0-C), it has a nominal focal length of 10 m ± 0.5% and the unobstructed diameter (clear aperture) was measured to be 46.3 mm. The curvature of the lens produces a linear deviation angle across the horizontal chords of the lens. This range of deviation angle is much greater than that produced by the wedged optics described previously. The BOS system is able to display a range of greyscale levels completely across the horizontal diameter of the lens, which for a 46.3 mm diameter optic corresponds to ±477 arc-seconds. For the schlieren images, the lens produces a range of deviation angles that is larger than the measurable range for both the standard and high sensitivity configurations. The middle image of Fig. 8 shows the range for the 3.03 mm source size schlieren system, and from [1] the range can be calculated to be ±123 arc-seconds, which corresponds to approximately the middle third of the lens image. For the rightmost schlieren image, the range is reduced by almost a factor of three by making the source only 1.12 mm in size, and thus the region of the lens over which there is intensity variations is only approximately the middle 1/9 of the lens and the range is ±46 arc-seconds from [1]. Of course the trade-off is a greater sensitivity of the schlieren system
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
using the narrow slit, by the same factor of three. This set of three images does demonstrate that the BOS technique, for some situations, does have a larger effective dynamic range than for a standard schlieren system. It is possible to use a graded filter to increase the range of the schlieren system but the sensitivity will be reduced [1, 4].
Fig. 8 BOS standard configuration image (left) and schlieren (centre, right) images of a 10 m focal length lens. The schlieren system standard sensitivity uses a 3.03 mm slit (centre), and high sensitivity with a 1.12 mm slit (right). 4. BOS and schlieren flow field imaging of a cone-cylinder model Both systems were used to image the compressible flow around a cone-cylinder model over a Mach number range from 0.6 to 1.2, and at two total pressures (60 kPa and 120 kPa). The conecylinder model was mounted on the TWT roll/pitch main model support arm and levelled in pitch. The model is a cylindrically-symmetric cone mounted on a cylinder with both the cylinder and the cone axes collinear, as shown in Fig. 9. The cylindrical body diameter is 33 mm and 256 mm long. The nose element is a 30° cone and 64 mm long.
ɸ 33
30°
256
64
Fig. 9 Dimensions (in mm) of cone-cylinder model. Figure 10 shows instantaneous schlieren images for nine Mach numbers. All images are for the same total pressure, Pt, of 120 kPa. The total temperature remained at 35°C ±2°C for the entire test. The standard schlieren system images have a larger field-of-view than the BOS images. They utilize the entire diameter of the windows (406 mm) throughout the test section, so their field-of-view is a 406 mm cylinder through the test section. The schlieren images show the expected progression of events as Mach number is increased: localised supersonic flow is observed around the shoulder (or base) of the cone with a localised shock just downstream. As the Mach number is increased this region grows and a bow shock appears for sonic and supersonic freestream conditions. Figure 11 shows standard BOS images for the same flow conditions as the schlieren images in Fig. 10. For each flow condition a set of BOS images was first obtained and then the BOS system
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
was rotated and translated out of the way of the standard schlieren system. In this way the BOSclassical schlieren system comparison was made rapidly and a comparison using test conditions on a steady set-point and ensuring minimal variability in test conditions. The BOS background diameter is 406 mm, and due to the conical perspective the field-of-view at the model is approximately 50% of 406 mm, or 203 mm. Another difference between the BOS and standard schlieren image sets is that the standard schlieren images show more spatial resolution than the BOS images. The schlieren images use the full 29 megapixels over a 406 mm diameter opening, whereas the resolution of the BOS image at the tunnel centerline is reduced in each direction because of the sampling size requirements of the image correlation technique required by BOS. This results in the spatial resolution of the BOS image being a factor of 1.9 worse than the standard schlieren system. The standard schlieren images show more detail over a larger area than the BOS images for this specific configuration of the BOS system. Overall a good qualitative agreement is obtained between the schlieren and BOS images. The exception is the flow at Mach 1.05, which will be investigated further in the next section. The schlieren images were obtained using 32 µs exposure time with the LED light source operated in the continuous mode, whereas the BOS images were obtained using the pulsed flashlamps whose nominal pulse duration was 5 µs. Both techniques freeze the motion of the shocks. It was determined in a previous section that both the BOS and schlieren systems have a minimum detectability that is approximately ±1 arc-second as-built for the TWT. The usefulness of minimum detectability is shown in Fig. 12, which is for a flow field that produces compressible flow features that are difficult to detect using either imaging technique. The figure shows airflow at a total pressure (60 kPa) that is half of the pressure used for the previous images, and as a result, the density gradients are half as much. Also, the 0.6 Mach number is the lowest Mach number that produces measurable density gradients at the shoulder of the cone. Both the BOS and the standard schlieren systems can detect the faint flow feature at the cone’s shoulder for 120 kPa. BOS successfully accomplishes this at Mach 0.6 and 60 kPa, whereas for both configurations of the schlieren system its minimum detectability is barely sufficient. The result corroborates with Section 3 which showed that the schlieren systems minimum detectability does not improve significantly for the high sensitivity configuration. Due to large f-stop (32) the BOS imaging near the object surface did not suffer from significant image blur as was reported by others in literature [3, 4]. This result could only be achieved with the powerful flash lamp light source providing adequate light intensity. More evident in the BOS images than the schlieren images are the effects of dust on the imager and blemishes or scratches on the schlieren windows. These result in erroneous data (white or black dots). These effects can be reduced by performing image processing on the corresponding locations and interpolating with the neighbouring pixel intensities.
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Fig. 10 Schlieren images (32 µs pulse length) using the standard sensitivity configuration, of the flow over the cone-cylinder model at Pt of 120 kPa and from M = 0.6 to 1.2. The flow is from right to left in all subsequent figures.
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Mach 0.95
Mach 0.85
Mach 0.6
Mach 1.0
Mach 0.98
Mach 1.1
Mach 1.05
Mach 1.01
Mach 1.2
Fig. 11 BOS Images (5 µs pulse length) using the standard sensitivity configuration of the flow over the cone-cylinder model at Pt of 120 kPa and from M = 0.6 to 1.2.
Mach 0.6 at 120kPa
Mach 0.6 at 60kPa
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
BOS
schlieren
BOS
schlieren high sensitivity
schlieren
Fig. 12 Standard sensivity configuration BOS (5 µs pulse length) and schlieren images (32 µs pulse length) of the flow at Mach number 0.6 at a Pt of 60 kPa and 120 kPa. High sensitivity configuration schlieren image shown for Pt of 60 kPa. 4.1 Effect of streamwise translation of BOS system As noted earlier, the BOS system has a conical perspective, while for the classical schlieren system the light rays traversing the test section are collimated. This can result in a distortion of the flow features that are being imaged by the BOS system with the effect being greater away from the optical axis of the BOS system. The distortion for the BOS system is shown dramatically in Fig. 13 for Mach 1.05 and Pt = 120 kPa, where a series of BOS images for different streamwise location of the BOS camera were obtained. The schlieren image, for the same flow shows what is expected: a bow shock near the cone tip and regions of locally supersonic flow around the cone base with corresponding shocks just downstream. The BOS images for the BOS camera at +80 mm and +140 mm show the reflections of the two flash lamps are evident in the images. These produce erroneous data upstream of the cone. In the BOS images, as the camera is moved from upstream to downstream, the bow shock (the most prominent compressible flow feature) can be seen to diminish and even disappear (BOS camera at -40 mm) in the images obtained at the downstream locations. In both BOS and schlieren the measured light ray refraction angle (and the density gradient) is a line-of-sight integral across the test section. The differences observed are due to the collimated versus conical perspective. With the difference expected to be greatest away from the BOS system optical axis. This is described more fully in Settles [1].
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
BOS camera -40 mm
BOS camera +40 mm
BOS original position
BOS camera +80 mm
schlieren
BOS camera +140 mm
Fig. 13 Instantaneous schlieren (32 µs pulse length) and BOS images (5 µs pulse length) of the flow over the cone-cylinder model at Mach number 1.05, Pt of 120 kPa for various streamwise positions of the BOS camera. Positive locations are upstream of the original position. 4.2 Schlieren calibration For the tunnel tests, the schlieren imaging system was calibrated using the same spherical PlanoConvex lens as in Section 3.3. Three methods of schlieren system calibration were performed. The first method was using the calibration lens and determining the reference intensity (Eref1) from the undisturbed background of a series of wind-off images (Fig.14a). The lens image pixel with intensity closest to that of Eref1 is the centre of the lens and from Eq.(3) the intensity can be plotted against refraction angle (Fig.14c). The linear region of data is selected and the experimentally determined sensitivity (or calibration) is obtained as the gradient of the line of best fit. The second method detects the edges of the lens and determines an averaged diameter, thus locating the geometrical centre, which corresponds to zero refraction angle. The intensity at this centre location is the reference intensity, Eref2 , and assumes a negligible loss of light intensity through the lens [4]. The difference between the geometric centre and the background intensity methods was found to be within 2 pixels (Fig.14b). The third method uses a region of the mean flow field image that is assumed to be of zero density gradient. For the quantitative comparison, Mach 0.95 at Pt of 120 kPa was analysed and a region upstream of the nose was sampled to provide the reference intensity, Eref3. For the standard sensitivity schlieren configuration the experimental sensitivity obtained from the first method (using Eref1) was 0.0074 contrast/arcsecond and 0.0075 contrast/arc-second for the second method (using Eref2), these values are 7.5% and 6.25% smaller than the theoretical sensitivity of 0.008 contrast/arc-second (obtained from
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Eq.2). For the third method, the effect of using the theoretical sensitivity and the sensitivity calculated from the image background (using Eref1) was assessed. (a)
(b)
(c)
Fig. 14 For that standard sensitivity configuration: (a) schlieren image of the calibration lens with sampled undisturbed regions used to calculate an averaged background reference intensity, Eref1, (b) close-up view of the lens – blue strip represents the locations where intensity measurements were sampled, (c) the extracted strip of intensity against refraction angle. 4.3 Quantitative comparison Applying a scaling calibration and appropriate transformation, the BOS and schlieren images were converted to contour plots of line-of-sight refraction angle (in arc-seconds). The refractive angle in a fluid flow is related to the density gradient by the Gladstone-Dale relationship [1, 2]. For axisymmetric or 2D flow, the density in the fluid can then be obtained by integrating the density gradient [2]. The refraction angles are presented here as it represents the raw results for comparison between the two techniques and without additional computation or integration required to obtain the density fields. A comparison between the refractive angle contour plots for both standard sensitivity configuration setups are shown in Fig. 15 for Mach 0.95 and a total tunnel pressure, Pt of 120
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
kPa. The averaged flow field presented was obtained from 50 images. As mentioned previously, the spatial resolution of the BOS measurements is 1.9 times lower than for the classical schlieren. Extracted profiles of the line-of-sight refraction angle at Y = 0.02 m from the centerline for both methods at Mach = 0.95 and Pt = 120 kPa are shown in Fig. 16. The schlieren system measures a lower peak refraction angle with the peak near the range limit for this configuration. According to schlieren theory [1], the range is ±123 arc-seconds, which is very close to what is obtained in Fig. 16. Whereas, BOS was able to measure a peak refraction angle of approximately 180 arcseconds at this location. The effects of the BOS conical perspective are evident, especially with features away from the camera lens centreline. The BOS profiles are generally stretched compared with schlieren, for example the double peak around X = -0.1 m is much wider as measured by BOS. The expansion fan minimum near the shoulder (X ~ -0.065 m) for both schlieren and BOS matches well (as it is near the optical axis). However, the schlieren system measures a double minimum whereas BOS only shows one. Figure 17 shows the line-of-sight refraction angle at Y = 0.054 m from the centerline for both methods at Mach = 0.95 and Pt = 120 kPa. The schlieren peak in Fig. 17 displays a top hat profile suggesting that the range limit of the system has been reached. Compared to Y = 0.02 m the BOS measurement at Y = 0.054 m shows a much larger peak due to the perspective view or the angle which the light penetrates the feature.
Fig. 15 Standard sensitivity configuration measured line-of-sight refraction angle of the flow over the cone-cylinder model at Pt of 120 kPa and Mach number of 0.95: schlieren (left) and BOS (right).
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
Fig. 16 Extracted schlieren and BOS line-of-sight refraction angle profiles of the flow over the cone-cylinder model at Pt of 120 kPa and M = 0.95, at Y = 0.02 m.
Fig. 17 Extracted schlieren and BOS line-of-sight refraction angle profiles of the flow over the cone-cylinder model at Pt of 120 kPa and M = 0.95, at Y = 0.054 m. 4.4 The effect of chosen reference intensity on calibrated schlieren Three methods of schlieren system calibration were attempted and the effect of each one on the schlieren results is shown Fig. 18. The processed schlieren results were found to be very sensitive to the reference intensity chosen, as can be seen in Fig. 18. Both the lens calibration methods, image background (Eref1) and the geometric centre of the lens (Eref2), showed an offset, particularly, evident upstream of the model where zero density gradient is expected (and measured using BOS). At positive refraction angles the offsets between the three methods is not as great. The method using the reference intensity upstream of the cone-cylinder model (Eref3) produced the near zero refraction angle as was measured with the BOS system. This suggests that quantitative schlieren can be performed (with reasonable accuracy) without undertaking a lens calibration. Provided, one can confidently assume that the region of the image chosen for the reference intensity has zero density gradient and the unobstructed source height is accurately measured to obtain the theoretical sensitivity. The difference between using the theoretical sensitivity and experimental derived sensitivity had a minimal effect on the results (green vs black in Fig. 18).
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
The lens calibration method proves to be problematic as it is done at the end or the start of a test, the temperature and pressure differences in the test section may affect the reference intensity reading and therefore the accuracy of the end result. Another factor is the effect of ground vibration or relative movement of components. The schlieren system is inherently sensitive to vibrations as any relative displacement between optical components causes changes in the optical path of the light and results in a different unobstructed source height. A different unobstructed source height produces an image background reference intensity that is either brighter or darker which changes the sensitivity of the system. The reference intensity obtained from the mean flow image performed the best as it takes into account the tunnel conditions and the average movement of components. To quantify all of these effects further investigation is warranted.
Fig. 18 Schlieren and BOS line-of-sight refraction angle of the flow over the cone-cylinder model at Pt = 120 kPa, M = 0.95, and Y = 0.02 m. Three different schlieren calibrations are shown. 5. Further improvements Quantitative schlieren relies on a constant light source and an even background intensity over the field-of-view. An even background intensity is obtained by accurate alignment and positioning of the schlieren system components and the use of high quality optics. Although the schlieren system mirrors produce reasonably high quality results they have an optical flatness of λ/2. The pixel resolution of each camera was the same but the pixel depth was not. The schlieren system would benefit from added pixel depth as well as a brighter light source which should result in higher sensitivity, better minimum detectability and a reduced measurement uncertainty. Further improvements in sensitivity can be obtained with longer focal length schlieren mirrors (but however this is not practical here considering the space constraints) and a larger focal length imaging lens which will allow higher spatial resolution, if one wants to ’zoom in’ on an area of interest. A possible reason for the differences between theoretical and experimental sensitivity is the error in measuring the unobstructed source height. Improvements in accuracy could be obtained by using a motorised stage. Movement of optical components or
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
vibrations can significantly affect the results, they can be difficult to isolate and effort should to be taken to reduce them as much as possible. A complete instantaneous in-situ calibration, including measurement of the unobstructed source height would be ideal. In the case of BOS, a larger field-of-view could be obtained by applying the background pattern to the entire wall not just the test section window. As with schlieren, higher spatial resolution can also be achieved with BOS by using a larger focal length objective lens to ‘zoom in’ on features of interest. By using a background pattern with continuously changing random dots one could obtain better accuracy and resolution by performing a sum of correlation technique over many images [8]. This approach only works for the averaged flow field. To avoid the perspective effects of BOS, Elsinga et al. [3] used collimated light projected through the background dot pattern and showed results with 2-3% measurement error. Both of these enhanced methods require an elaborate background setup which is not always possible in a production facility. 6. Conclusions The performance of a standard schlieren imaging system was compared with a BOS imaging system in a large-scale transonic facility. By using similar cameras, with the same geometric constraints and test conditions, a measure of the relative performance was obtained. The two tested configurations, standard and high sensitivity, produced the expected increase in sensitivity, however there were discrepancies in the absolute values compared to theory. The experimentally derived minimum detectability was found to be similar and of the order of ±1 arc-seconds for both systems. Both systems were used to image the compressible flow around a cone-cylinder model, over a Mach number range from 0.6 to 1.2. Both systems could discern the weakest flow feature created around the test article, which was the expanded flow around the shoulder of the cone at Mach 0.6. The effect of the BOS system conical perspective view was evident, by translating the BOS system in the streamwise direction, at 1.05 Mach, a relatively strong bow shock at the tip of the cone-cylinder model could be made to disappear from the BOS image. When comparing extracted profiles of refraction angle, at 0.95 Mach, a good match between schlieren and BOS was obtained near the camera lens centerline. The BOS conical perspective was again evident, especially away from the lens centerline where the BOS profiles appeared stretched compared with schlieren. The BOS system’s dynamic range was shown to be significantly greater than that of the standard schlieren system. Regions of the schlieren image were over ranged resulting in a loss of information. The effect of the chosen reference intensity in performing calibrated schlieren was assessed. The reference intensity chosen from a mean flow image of a region of the flow assumed to have near zero density gradient performed best, whereas the reference intensity obtained at the start or the end of the test showed an offset. The discrepancy was mainly attributed to differences in the tunnel conditions and movement of optical components. In summary, for the tested configuration in this study, BOS performs as well as schlieren in terms of minimum detectability and has a larger dynamic range but schlieren has higher spatial
18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics・LISBON | PORTUGAL ・JULY 4 – 7, 2016
resolution and doesn’t suffer from the perspective view experienced by BOS. BOS provides both the horizontal and vertical density gradients whereas classical schlieren only provides the gradients normal to the knife edge. The other advantage of the BOS technique over that of the classical schlieren is that it can be used in situations with a limited field of view to the test article, where a direct light path though the test section is not available, and that it does not require expensive optical quality glass elements. 7. Acknowledgements The authors would like to thank Jamieson Kaiser for his contributions to the investigation. Many thanks also to Peter Macaluso from AEDC and the DST Group wind tunnel support crew, Paul Jacquemin, Kevin Desmond and Peter O’Connor for assistance in conducting the test. 8. References [1]
Settles G (2001) Schlieren and Shadowgraph Techniques: Visualizing Phenomena in Transparent Media. Springer-Verlag.
[2]
Raffel M (2015) Review Article: Background-oriented schlieren (BOS) techniques. Exp. Fluids 56:60.
[3]
Elsinga G, Van Oudheusden B, Scarano F and Watt, D (2004) Assessment and application of quantitative schlieren methods with bi-directional sensitivity: calibrated color schlieren and background oriented schlieren. Exp. Fluids 36:309-325.
[4]
Hargather M and Settles G (2012) A comparison of three quantitative schlieren techniques. Optics and Lasers in Engineering 50:8-17.
[5]
Aero Systems Engineering (2000) AMRL - Transonic Wind Tunnel Users Guide.
[6]
Giacobello M, Manovski P, Lam S, Premachandran S, Kleine H (2011) Application of Monochrome and Colour Schlieren to the Study of the Flow Past a Cone-Cylinder at Transonic Speeds, 9th Australasian Heat and Mass Transfer Conference, Monash University, Melbourne, Australia.
[7]
Raffel M, Willert C, Wereley S, Kompenhans J (2007) Particle Image Velocimetry – A Practical Guide. Springer-Verlag.
[8]
Schröder A, Over B, Geisler R, Bulit A, Schwane R, Kompenhans J (2009) Measurements of density fields in micro nozzle plumes in vacuum by using an enhanced tomographic Background Oriented Schlieren (BOS) technique. 9th International symposium on measurement technology and intelligent instruments, Saint-Petersburg.