Advances in the Phase Doppler Method for Dense Spray Measurements

Transcription

1 Advances in the Phase Doppler Method for Dense Spray Measurements W. D. Bachalo*, G. A. Payne, C. M. Sipperley, K.M. Ibrahim Artium Technologies, Inc. Sunnyvale, California, USA Abstract Phase Doppler Interferometry (PDI, PDA, PDPA) continues to be one of the most useful spray measurement instruments since its invention in A key limitation to PDI has been the requirement that only one droplet passes the measurement volume at a given time. This limits the maximum particle number density that the instrument can reliably measure, based on random arrivals. Coincident particle occurrences also introduce the probability of underreporting values such as number density, volume flux, and volume-based statistics. In this paper we present a new approach to capturing and analyzing PDI signals that allows for multiple droplets in the measurement volume at the same time. Existing signal processing systems rely on analog and/or digital Doppler burst signal detection that are reliable when coincident arrivals of particles at the probe volume are limited. However, under high number density conditions, these signal detection systems can remain active during several particle arrivals and accept this information as a single event only to be rejected later or to produce a faulty measurement. A new signal processing approach has been developed that allows for capturing long burst records that can contain multiple Doppler signals. New processing algorithms use this extended record and partition signals into sub-records which are then processed for the individual droplet signals. This approach enables recovery of individual measurements that would have been lost with conventional phase Doppler signal processors. Results when measuring a monodisperse droplet stream which has a known droplet generation rate with a large measurement volume (guaranteeing multiple droplets at the same time) are presented to show the method can recover the full data rate in contrast with conventional systems. Additional results from measurements in a dense spray are compared with a simple sampling tube to show recovery of the volume flux and number density in challenging environments.

2 Introduction Since its invention originating in 1980, the phase Doppler method has proven to be a valuable instrument for characterizing sprays and spray droplet dynamics over a wide range of conditions and including complex turbulent two-phase flows and spray flows with reaction. The method has benefited from extensive development over the past decades. Advancements in the theory of light scattering interferometry have helped to reveal characteristics of the method that required further refinement and advancement in the methodology. Such problems as particle trajectory errors and mixed mode light scattering have been addressed and innovations have been applied to cope with these potential sources of measurement uncertainty. Under normal spray conditions, the method has been demonstrated as being reliable and produces accurate spray droplet size and velocity measurements. With innovations used to characterize the dimensions of the measurement volume in situ, the measurement of droplet number density, liquid volume flux, and liquid water content have been shown to be accurate under most conditions. Remaining challenges for the instrument involve the need to measure dense high-speed sprays associated with modern diesel, gasoline direct injection and gas turbine combustor sprays. These sprays are typified by very high number densities and high droplet velocities with bulk velocities that reach or exceed supersonic. Added to the difficulties are the high temperature and high pressure conditions of these environments. To contain these spray flows, thick windows are necessary. In some cases, the measurements are attempted in optical engines with thick cylindrical windows which further exacerbate the measurement challenges. Although measurements have been made under such conditions, one might suspect the reliability and accuracy of these results given the very challenging measurement conditions. Phase Doppler interferometry (PDI, PDA, PDPA) instruments are single particle counter devices which require a high probability of only one particle residing in the measurement volume at one time. Prior to the papers by Bachalo [1] and Bachalo and Houser [2], single particle light scattering methods used confocal light scatter detection which was deemed necessary by researchers working in particle and spray droplet sizing. In his work, it was revealed that large off axis light scattering angles could be used to size particles with the significant advantage of being able to reduce the size of the measurement volume by orders of magnitude. The change was significant in coping with spray measurements which by current standards would not be considered dense sprays. Furthermore, with off axis light scatter detection, apertures with widths as small as 15 µm can be used to limit the extent along the laser beam that is observed for the measurements. A very small measurement volume is formed that can, in fact, be made as small as the largest drops to be measured or smaller. The laser beam diameter can also be focused to a diameter consistent with the droplet size range to be measured and thereby further reducing the size of the measurement volume. However, as will be explained in the following sections, there are limitations to these parameters dictated by the light scattering mechanisms, spray flow conditions, and the droplet sizes to be measured. In this paper, we will review the conditions for mitigating coincident errors and trajectory dependent sizing errors under dense spray conditions. Some of the methods have been disclosed previously but will be reviewed here for convenience and to set the current innovative approach in proper perspective. Limitations on minimizing the sample volume size and the probability of coincident occurrences will be specified in detail. Examples of coincident events will be provided along with the response of the instrument to these occurrences. A new method involving the detailed recording of signals wherein several particles pass the sample volume in succession or that are partially overlapped and parsing of these signals will be described. Examples showing the benefits of this approach are provided and discussed. Experimental Methods Challenges in Dense Spray Measurements The description of dense sprays is often ambiguous and stated without a specific definition. In this work, dense spray does not represent the spray formation region wherein the liquid is undergoing breakup with droplets not yet fully formed. Since the light scattering method is dependent upon the droplets being spherical or quasi-spherical, the descriptions and comments herein describe the region where the spray droplets are fully formed. In an attempt to define high number density sprays, the following characterization is provided: S 10 (1) D where S is the mean separation distance between particles and D10 is the arithmetic mean diameter. Although this expression may be used as a guide, it is not very useful if one cannot estimate the mean separation distance S because the number density is too great. Nonetheless, it provides a better description 10

3 P(n) ILASS-Americas 29th Annual Conference on Liquid Atomization and Spray Systems, Atlanta, GA, May 2017 than is normally offered in presentations discussing dense sprays. An estimation of the probability of having coincident measurements (more than one particle in the sample volume at one time) may be provided using Poisson statistics. The probability of finding n particles within the probe volume V can be expressed as follows: n ( VN ) ( VN ) P( n) e (2) n! where N is the particle number density. Using this expression, the probability of finding only a single particle (n = 1) within the probe volume can be expressed as: ( P( ) ( VN) e VN ) 1 (3) The variation of P(n) is shown in figure 1. One might assume that the probe volume, number density product VN providing the maximum probability P(1) of one particle in the probe volume would represent the optimum condition. However, as seen in figure 1, the probability of two particles, P(2) in the probe volume is significant at that condition (VN product) indicating the probability of coincidence is relatively high. The goal is to maximize the P(1)/P(2) ratio to achieve an acceptably low probability of coincidence. From this plot, it is evident that even with a very small VN product, there remains a probability of coincidence. It is well-known that coincident measurements can lead to measurement error which will be discussed in subsequent sections P(1) P(2) P(3) P(0) VN Figure 1: Probability for coincidence based on Poisson statistics. When challenged by high spray number density N, the obvious strategy to enable measurement of the spray condition is to reduce the size of the sample volume V. The sample volume is defined by the square of the effective focused laser beam diameter at the laser beam intersection of the PDI instrument and the size of the slit aperture projected from the receiver across the focused laser beams as well as the offset collection angle. There are limitations on the size of the focused laser beams imposed by the condition of trajectory dependent light scattering, Sankar, et al., [3] and [4]. With the random trajectories through the Gaussian beam intensity distribution and the sample volume, there is a significant probability that the glare point for reflection from the drops will pass the beam with a maximum incident intensity whereas the refracted glare point passes closer to the edge of the Gaussian beam. This results in light scattering by the reflection and refraction components at similar orders of magnitude with the consequence of producing a complex non-sinusoidal interference fringe pattern at the receiver plane. Light scattering trajectory errors can result in the instrument reporting spurious large droplet measurements. In an effort to mitigate this problem, some researchers have elected to use light scatter detection angles at approximately 72 which is near the Brewster s angle for the incident light and hence, the light scattered by reflection is minimized. We have generally regarded this approach as unacceptable for two reasons. The light scattered intensity is approximately eight times less at 72 than it is at 40. Furthermore, for fuel sprays or other liquids with index refraction greater than approximately 1.4, the light scattering by internal reflection, p = 3 is significant at this angle. This defeats any advantage to using this light scatter detection angle. An additional constraint on the focused beam diameter is the issue of transit time and its effect on the signal processing. The transit time given as: D/ v (4) where D is the 1/e 2 diameter of the focused laser beams and v is the particle velocity. As a result of the Gaussian light beam intensity, it is well-known that the smaller particles must pass through the focused laser beam closer to the peak intensity to produce a detectable signal. Hence, the effective diameter of the sample volume for the smallest particles is significantly smaller than the specified laser beam diameter. For high velocities associated with gas turbine and diesel sprays, for example, the droplet velocities may exceed 100 m/s. Although claims have been made regarding the ability to process signals that have duration much less than one microsecond, these claims are based on ideal conditions with high signalto-noise ratio. We have demonstrated that the particle transit time has a direct effect on the variance in the phase and frequency measurements used to infer droplet size and velocity. For example the following expression shows the effects of the various signal parameters on the measurement uncertainty, Ibrahim, et al., [5]:

4 var f 12 (2 ) T B(1 1/ N) SNR 2 3 r (5) where f is the signal frequency, T r is the measurement time which may be less than or equal to the particle transit time, B is the signal bandwidth, N is the number of samples in the ADC record, and SNR is the signalto-noise ratio. Clearly, if the signal-to-noise ratio is low, short particle observation times will result in significant frequency variance and similarly, variance in the phase measurement used for particle velocity and size estimations. Under dense spray conditions, the signal-to-noise ratio will deteriorate due to multiple light scattering, beam degradation due to droplets passing the laser beam close to the measurement volume, and light beam attenuation. Thus, the focused laser beam diameter is limited by the particle speed and consequently, the transit time. The slit aperture in the receiver provides an additional means to limit the size of the measurement volume. Using a small slit aperture limits the length along the laser beams over which particles are observed and also blocks multiply scattered light from the proximity of the probe volume. The current phase Doppler instruments generally offer a number of slit apertures that can be selected to minimize the measurement volume size when attempting to measure in dense spray environments. However, there are also limitations on the size of the slit aperture that may be used. When measuring high pressure sprays through thick windows, the image quality can be expected to be degraded by optical aberrations. Hence, very small slit apertures may reduce the observed scattered light to an unacceptable level resulting in low signal-to-noise ratio while failing to reduce the volume over which droplets may be detected. Furthermore, when dealing with multi-hole pulsed injectors, gas turbine spray flows, and other sprays wherein the droplet angles of trajectory may vary significantly, a small slit aperture can result in unacceptably low signal observation times. When the focused beam diameter is set to a value that provides adequate signal observation time, droplets with trajectories not aligned with the slit aperture will result in significantly smaller sample observation times and may fail to produce validated signals. This condition has generally been overlooked as a potential problem in the application of phase Doppler instruments but can generate consequential measurement bias and uncertainty. Our PDI instruments have been designed to allow rotation of the slit aperture into the main particle angle of trajectory but this only works if the droplets are moving in a relatively narrow range of angles of trajectory. Multiple particle light scattering analysis Coincident occurrences (more than one particle passing the sample volume at one time) are often assumed to produce individual Doppler burst signals that may be partially overlapped. This assumption is not entirely correct since the light scattering is a coherent interaction of the light fields scattered by the different particles from each of the two laser beams. For two or more particles, all the scattered light fields interfere to form a resultant complex interference fringe pattern which will be non-sinusoidal. Our earlier work (Sankar, et al. [6]) described this light scattering mechanisms in detail using an advanced geometrical optics approach. We have updated our analysis using a version of Mie theory. (Note: This is strictly not Mie theory since Mie theory describes the light scattered by an isolated homogeneous sphere in a homogeneous environment illuminated by a monochromatic plane light wave. Dense sprays violate the isolated sphere requirement.) Far-field scattering of laser light by two or more droplets may be approximated as multiple, superposed Mie scattering fields. The authors have developed such a simulation package to predict the integral intensity across the detector areas of a phase Doppler instrument. The light scattered by each laser beamdroplet combination is computed independently. A grid covering the sensitive area of the receiver face for each detector is generated. For each grid location the far-field intensity and relative phase of the Mie Scattering is calculated for each direction component utilizing the methodology of Bohren and Huffman, [7] as implemented by Laven,[8]. To calculate the phase shift between detectors for a single drop, one need only step through a phase shift on one of the beam pairs before summing the fields with phasor summation. To simulate real drop signals the intensity of the initial beam must be considered as a function of time (e.g. a Gaussian beam profile) and a total frequency chosen (Bragg cell driver plus Doppler difference). The authors' simulation package allows any number of droplets each with variable intensity profiles, frequencies, and relative phases. The intensity profiles for each droplet encompass relative intensities as well as variable stop and start times. For each time step through the simulation, the total intensity at each point in the grid is determined from a phasor summation of the scattering contributions from each beam striking each drop. The total intensity is integrated to determine the time-varying signal that would be recorded by a phase Doppler instrument for that detector. The result of one such simulation is shown in Figure 2. In this figure, a 10 micron droplet is followed by a 30 micron drop with the same frequency,

5 intensity, and crossing time. Also note the difference in how signal 'C' lags 'A' between the left and right plots. The larger drop (right) has a larger phase shift between detectors as would be expected for a phase Doppler signal. Comments on Signal Processing Signal processing based on the discrete Fourier transform provides the optimum means to accurately detect and process signals to obtain frequency and phase information as has been demonstrated by Sankar, et al. [6]. A detailed analysis of the requirements for frequency and phase measurements of Doppler burst signals is presented in Ibrahim and Bachalo, [9]. In that paper, it was clearly shown that quadrature (complex) sampling of the signals provides the greatest accuracy and reliability when measuring frequency and phase under typical spray conditions. With this approach, the signal is represented as s( t) r( t) iq( t) (6) Figure 2: Simulated Doppler signal resulting from the coherent interaction between two droplet signals for a 10 m and a 30 m with the second arriving after the first droplet has entered its measurement probe volume. The top plot is the simulated signal for the center detector (B). The bottom three plots are 100 ns subplots for the outside detectors: before the second drop, centered between drops, and after the first. Digital Signal Detection Signal processors for LDV and PDI applications require signal burst detection systems to sample the random arrivals of particles that produce Doppler signals. These detection systems must be capable of reliably detecting signals over a very wide dynamic amplitude range and should not produce significant false detections due to noise. Ibrahim and Bachalo, [9] introduced a method for signal burst detection based on SNR using a real time Fourier transform approach. This method facilitated detection of Doppler signals with SNR down to 0 db or lower without the need for user intervention, as was the case for the existing where i is the imaginary number (i 2 =-1) and r( t) Acos 2 f t n( t) (7) 1 q( t) Asin 2 f t ( t) (8) where n is the Hilbert transform of n(t). When sampled, the digital form of the equation to obtain signal phase is given as: K r( i)cos2 iks/ Nq( i)sin2 iks/ N 1 i0 (9) tan K r( i)sin2 iks/ Nq( i)cos2 iks/ N i 0 where r(i) and q(i) represent the in-phase (real) and the quadrature (imaginary) components of the sampled signal s(i), N is the number of samples used for phase measurement, and k s is the signal frequency measured using the DFT algorithm. Acquiring complex sampled signals is clearly superior when estimating the phase of the sampled signal and the resolution depends on the number of samples, N. analog threshold voltage level detection methods. More recently, Bachalo, et al. [10] and [11] introduced an advanced method for digital detection which allows the use of 32 or more complex digital samples for signal detection. The signal detection is based on the continuous digital sampling of the analog signals from phase Doppler detectors. These signals are sampled in quadrature and the quadrature signals are continuously analyzed for signal-to-noise ratio. When a coherent signal is detected (SNR rises above a predetermined level), a burst detection gate is set (green square wave curve in figure 3) and that portion of the record is transferred to the computer for higher level processing using the full complex discrete Fourier transform (DFT). This method has proven to be even more reliable than the previous digital approach in detecting signals under low SNR conditions and especially for dense sprays. It is important to recognize that for sprays with a size range of 33 to 1, the signal amplitude range is approximately 1000 to 1 so the smallest droplets will produce very small signals with low SNR. Examples of the signal detection under normal and dense spray conditions are shown in figures 3. These examples show the efficacy of the burst detection system in locating Doppler burst signals and separating them even when the signals are nearly contiguous as shown in the first frame. In the center frame, two Doppler burst signals are coincident which produces an overlapping signal. This condition is detected as a single event since the filtered signal remains coherent over the duration of both signals. Once processed using full complex Fourier 1 n

6 transforms, the signal may be rejected if the phase shift and frequency are not consistent over the entire signal duration. In the right frame, three signals arrive in quick succession with the burst detection system remaining active over the three signals. Once again, with conventional signal processing, these three signals would be analyzed using the full complex Fourier transform and rejected because of inconsistency in frequency and phase over the duration of the detected signal. Under very dense spray conditions, detection of multiple particles as a single event will lead to a high degree of signal rejection and sampling bias towards the larger particles. Often, these conditions occur without the knowledge of the experimenter and the measurements are reported as valid. Clearly, even the most advanced signal detection means can be defeated under dense spray conditions. Figure 3: Three examples of signal detection for dense sprays. Yellow (top) trace is the raw signals, the magenta (middle) trace is the down-mixed, high pass filtered, log amplified signal, and the green (bottom) trace is the signal detection. Parsing signals Our current signal processing strategy is to sample the signals detected by the burst detection system in quadrature and pass the digital replicas to the signal processing means. The PDI signals consist of complex sampled (real and imaginary) records for each of the three signals for each droplet. The sampled record may have lengths consisting of up to 100,000 ADC samples for each detection. This allows the acquisition of digital records over multiple signal events which may be many particle signals during a single detection. For example, in figure 3, detections for two and three signals are recorded as a single event and passed to the computer for processing. If processed as a single event, the validation logic may either reject the signals, report a measurement for only the largest particle, or it may report a faulty size and velocity measurement. To avoid such errors, we have implemented a parsing strategy to separate the event provided by the burst detector into individual particle signals. Signal parsing is implemented by iteratively reprocessing the sampled digital replica of the detected event using a number of strategies. A high quantization (12 bit ADC) sampler is used to obtain an accurate amplitude record of the event under the burst detector. As seen in figure 3, the individual Gaussian signal peaks and shapes are easily discernible and this information can be used as a first effort to separate the individual signals under the burst detection gate. By iteratively applying the Fourier transform to sample segments of the signal, changes in signal to noise ratio, frequency, and phase of the signal segments are used to further parse the signals into their individual Doppler bursts. Based on the estimated particle velocity, the expected signal duration is also applied in the separation algorithm. Since the signals are recorded with high resolution (large number of samples, N), segments may be evaluated at varying lengths (number of samples N i). Modern computers with eight or more cores are very fast so parsing of the signal can be done in near real time. The computational effort required will depend upon the dense spray characteristics which determines the number of iterations required to accurately and reliably parse the detected events into the individual signals and to extract their frequency and phase. For example, under diesel and direct injection gasoline sprays, a significant number of signals may be detected as a long single event by the burst detector. In such cases, the parsing requires a number of iterations to extract the individual Doppler signal information.

7 Another application of this approach is to determine the frequency change of a particle as it transits the sample volume. Under relatively dilute particle field conditions, this approach can be used to assess local particle accelerations in turbulence. The signal burst duration can be separated into any number of segments and each segment processed using high-resolution DFT s to obtain instantaneous frequency (velocity) changes of the particle as it transits the measurement volume. We are also investigating the possibility of seeding a spray flow with small seed particles (~2-5 m) at a high number density and measuring these particles in close proximity to droplets. Such dense seeding will result in signal coincidence that can be reconciled with our parsing method. This approach would provide better information on the particle Reynolds number and local slip velocity which affects particle drag, evaporation, deformation and breakup. Results and Discussion Experimentation has been conducted to evaluate the method of parsing signals to determine how well coincident signal events can be separated and whether this approach can significantly improve the validation rate, eliminate false large particle measurements, and improve on droplet number density and volume flux estimations. A standard modular PDI instrument (Artium MD 200) was used in these experiments. The sample volume was set to be larger than normal to exacerbate the problem of coincident measurements and multiple particle detections. In the first experiments, a monodisperse droplet generator (MDG) was used to generate a stream of droplets very close together with predictable coincidence. Since the drop size and arrival rate are known to high precision and accuracy, this basic test allowed evaluation of the parsing scheme and assessment of the coherent light scattering effects from one droplet upon the other. An example of the monodisperse droplet stream signals are shown in figure 4. Variable degrees of signal overlap were tested. One might argue that spray droplets will pass on random trajectories through the sample volume. However, based on the requirements for adequate transit time and trajectory dependent light scattering, the focused laser beam diameter is limited to a minimum size. The slit aperture in the receiver can be as small as required or as limited by the diffraction resolution limit of the receiver lens. This results in a long slender sample volume geometry and consequently, a high probability of drops being detected as serial transits more often than parallel. Figure 4: Examples of monodisperse droplet signals for single droplet and multiple droplet detections. The upper frames show the three signals from the three detectors and the green trace is the gate signal for droplet detection. The lower frames show the raw signal, the filtered, downmixed, and log amplified signal for single and multiple particle detections. The droplet rate based on the excitation frequency of the MDG was 6890 and the sampling rate was also 6890 for the noncoincident condition. Under the simulated dense spray condition, the sampling rate dropped to Using the reprocessed parsing approach, the data rate was recovered to be Although not a perfect recovery, the improvement was significant. It was interesting to note that the mean diameter, D10, did not change by more than 1 m. This is due to the fact that all droplets in the string of droplets passing the probe volume had the same size and produce a similar result whether coincident or not. Moderately dense spray measurements were obtained using a larger than normal sample volume to produce a significant probability of particle coincidence. Flux was also measured with a sampling probe over an extended period of time and these measurements were used for the flux comparisons. PDI spray measurement results were then parsed using the approach described above to obtain a refined characterization of the spray. Measurements of droplet size, number density and volume flux were recorded. The measurements were then repeated using a more highly focused laser beam which minimized the probability for coincidence. Results were then compared to assess the efficacy of our signal parsing scheme to obtain dense spray measurements, table 1. For this moderately dense spray with approximately 15% coincidence, the effect on droplet size and number density (counts) was not very high. However, there is a significant improvement in the flux measurements. Future work will address dense sprays associated with diesel and gasoline direct injection sprays. The impact on these spray measurements will be much more pronounced since these sprays will have a very high rate of coincidence.

8 Summary and Conclusions Measuring dense sprays with any optical method can be challenging due to laser beam attenuation, multiple light scattering, and particle coincident occurrences in the sample volume. Although measurements have been reported in high number density sprays, little has been done to verify the claims of measurement reliability and accuracy. Basic Poisson statistical analysis can be used to estimate the probability of coincidence but without knowing the spray droplet number density a priori, these results can only be used as a rough guide. In dense spray environments, it is very unlikely that reliable number density measurements can be made. In this paper, we reviewed the measurement conditions and discussed the parameters and limitations that affect the ability of the phase Doppler method when measuring high density sprays. We emphasized that the term dense sprays does not include the spray formation region and we also offered a more rigorous description of a dense spray. An advanced method for mitigating the measurement errors that may be expected when measuring dense sprays was introduced. Because the signals may be arriving in very close proximity and will partially overlap in the measurement volume, we have developed a means for digitally recording replicas of the signals and the parsing these signals using our software algorithms to separate the signals produced by individual particles. Our approach has been shown to provide a significant improvement in the reliability and accuracy of the spray measurements. Basic tests on monodisperse drop streams and dense sprays were used to quantify the improvements realized. Acknowledgements We are grateful for the support of this work by the U.S. Army PEO STRI under the SBIR program. References [1] Bachalo, W.D. Appl. Opt (1980). [2] Bachalo, W.D. and Houser, M.J. Opt. Eng. 23, 1984, [3] Sankar, S.V. and Bachalo, W.D. Applied Optics, April 20th, Volume 30, #12 (1991). [4] Sankar, S.V., Inenaga, A.S., & Bachalo, W.D., Proc. 6th Intl. Symp. on the Appl. of Laser Techniques to Fluid Mechanics, Lisbon, Portugal (1992). [5] Ibrahim, K., Werthimer, D., and Bachalo, W.D., Proceedings of the ASME Fourth International Conference on Laser Anemometry, Ohio, pp (1991). [6] Sankar, S.V. and Bachalo, W.D., Proc. 7th Intl. Symp. on the Appl. of Laser Techniques to Fluid Mechanics, Lisbon, Portugal, (1994). [7] Bohren, C.F. Huffman, D.R., Absorption and Scattering of Light be Small Particles, New York, Wiley Press, (1983). [8] Laven, P., MiePlot v4221s, (2010). [9] Ibrahim, K., and Bachalo, W.D, Proceedings of the 8 th International Symposium on Applications of Laser Techniques to Fluid Mechanics, Lisbon, Portugal, (1996). [10] Bachalo, W.D., and Ibrahim, K.M., U.S. Patent 5,289,391, Feb. (1994). [11] Bachalo, W.D., Ibrahim, K.M., and Payne, G.A., U.S. Patent 7,564,564, July, (2009). Further testing is required to fully develop and evaluate this innovative approach. Tests will be conducted on diesel and gasoline direct injection sprays. Since we save the digitized signals, it is easy to re-process the signals and compare the results using these different strategies. We are also interested in observing small scale accelerations in turbulent flows and in estimating local slip velocities. Saving long high precision recordings of the signals allows observation of the particle velocity over small segments as the particle transits the measurement volume. These results will be of interest in turbulent and two-phase turbulent spray flow experiments.

9 Gain, volts Counts PVC Counts RP PVC D10 D10 RP Flux Flux Diff. % Flux RP Flux RP Diff. MVD MVD RP Table 1. Results for a moderately dense spray (approximately 15% coincidence showing the effect of parsing the signals (RP). The improvement in the flux measurement is significant.