NASA Technical Reports Server (NTRS) 19980019288: Accuracy Study of a 2-Component Point Doppler Velocimeter (PDV)

12 Pages·0.82 MB·English

by  NASA Technical Reports Server (NTRS)

207325
0 C,, 7"
AIAA 97-1916
Accuracy Study of A 2-Component Point
Doppler Velocimeter (PDV)
J. Kuhlman, S. Naylor, K. James and
S. Ramanath
West Virginia University
Mechanical and Aerospace Engineering Dept.
Morgantown, WV
28th AIAA Fluid Dynamics Conference
4th AIAA Shear Flow Control Conference
June 29 - July 2, 1997
Snowmass Village, CO
For permission to copy or republish, contact the American Institute of Aeronautics and Astronautics
1801 Alexander Bell Drive, Suite 500, Reston, VA 20191- 4344
AIAA-97-1916
ACCURACY STUDY OF A 2-COMPONENT POINT
DOPPLER VELOCIM TER (PDV)
John Kuhlman _, Steve Naylor_,Kelly James2,and Senthil Ramanath 2
West Virginia University
Mechanical and Aerospace Engineering Department
Morgantown, WV 26506-6106
[phone]
ABSTRACT intnmve, planar imaging, Doppler-based
velocimetry technique, as well as the accuracy of
A two-component Point Doppler Velocimeter related Point Doppler Velocimetry (PDV). Both of
(PDV) which has recently been developed is these techniques use an iodine vapor cell absorption
described, and a series of velocity measurements line filter (ALF) to determine the Doppler shift, and
which have been obtained to quantify the accuracy of hence the velocity, of small seed particles in a flow
the PDV system are summarized. This PDV system field, as these panicles pass through a two-
uses molecular iodine vapor cells as frequency dimensional sheet of laser light. The same portion
discriminating filters to determine the Doppler shift of the light sheet isviewed through a beam splitter,
of laser light which is scattered off of seed particles either by a pair of video cameras (for DGV), or a
in a flow. The majority of results which have been pair of photodetectors (for PDV), with the iodine ceU
obtained to date are for the mean velocity of a ALF placed inthe optical path ofone of the cameras
rotating wheel, althoughpreliminaryd,ata are or photodetectors. Laser wavelength and ALF
described for fully-developed turbulent pipe flow. absorption band are matched such that the range of
flow velocities of interest yields Doppler shifted
Accuracyofthepresentwheelvelocitdyatais frequencies which lie in the linear portion of the
approximatel+yi% offullscalew,hileImcarityof absorption band of the ALF. As a result, the ratio of
asinglcehannelisontheorderof +-0.5% (ie+,0.6 the light intensities seen by the two detectors at a
m/see and ± 0.3 m/see, out of 57 m/see, point in the flow yields a signal which is
respec_ely). The observed linearity of these results proportional to the particle velocity.
is on the order of the accuracy towhich the speed of
the rotating wheel has been set for individual data For a non-scanned "point" PDV system, very
readings. The absolute accuracy of the rotating high data rates are possible, limited primarily by
wheel data isshown tobe consistent with the level of seeding/signsatlrengtahndA/D conversiosnpeed.
repeatability ofthe cell calibrations. Accuracyof sucha point system may be comparable
to that of conventional laser velocimetry (LV). Use
The preliminar/turbtdem pipe flow data chow ofconventional CCD camerastoview a region of the
consistent turbtdence intensity values, and mean lightsheetyields velocity dataatatypicarlesolution
axial velocity profiles generally agree with pitot of 640 pixels by 512 lines, at framing rates of up to
probe data. However, them is at present an offset standarvdideoratesof 30framespersecond, butata
errorin the radial velocity which is on the order of reducedaccuracy(typicalolnytheorderofabout+ 5
5-i0 % ofthe mean axial velocity. to 8 %). This reduced accuracy is primarily due to
camera noise; cooled cameras can reduce this error,
INTRODUCTION but at a significant increase in cost. Also, fi'ammg
rates are typically reduced to on the order of I Hz
Thisresearcphrojecitsexplorintgheaccuracy for these more accurate, cooled cameras.
of Doppler GlobalVelocimetry(DGV), a non-
A two-component non-scannedpointPDV
Copyright @1997 by the American Institute of system has beendevelopedtodate inthe current
Aeronautics and Astronaugcs, Inc. All rights project.A two-componentscannedDGV system
reserved. using CCD cameras is currentlyalso under
_Professor, Associate FellowAIAA development.The accuracylimitsofbothsystems
2Graduate Research Assistant arebeingsystematicalelxyploredt,hroughaserieosf
meas_ements in relatively simple, unheated flows out all signal butthe Doppler shifted frequencies due
such as fully-developed turbulent pipe flow, a to molecular Rayleigh scattering.
turbulent circular jet, and grid turbulence. A
rotating wheel is also being used as a velocity A preliminary study has been conducted by.
standard. The present paper describes the two- Hoffenberg and Sullivan (1993) to measure the
component PDV system, and presents typical Doppler frequency, and hence velocity, at a point
velocity measurements which have been obtained for using a non-scanned filtered particle scattenng
a rotating wheel, to assess the accuracy of the PDV (H'S) technique. Measuremem accuracy of both
system for mean velocity measurements. ALso, mean and turbulence quantities was comparable to
preliminary PDV measurements in the fully- conventional LV data at the exit of an axis3amnetric
developed pipe flow are presented. jet at about I00 ft/sec. However, large errors in
mean and RMS velocities were found near the edges
LITERATURE SUMMARY of the jet, possibly due to uneven seeding and low
signal-to-noise ratio. Similar point DGV studies
Several different non-intrusive whole field have been conducted by Morrisen et al (1994), and
velocimetry techniques are cunently under by Rcehle and Schodl (1994). Roehle and Schodl
development which provide velocity data in a plane, have improved the accuracy of their measurements
which can thus greatly reduce the time required to through active stabilization of the frequency of their
map out a complex flow field. It is expected that this CW Argon ion laser.
can lead to enhanced insight into flow physics. Of
these techniques, particle image velocimelzy (PlV) Recently, others at NASA Ames (McKenzie,
has perhaps been the most fully developed (Adrian I995), NASA Langley (Smith and Northarn, 1995),
and Yao, 1983). Scalar imaging velocimetry (SIV) aad Ohio State University (Elliott, et al, 1994, and
shows promise for determination of three Clancy and Samimy, 1997) have also developed
dimensional velocity data m large Schmidt number scanned DGV systems. Both Smith and McKenzie
liquid flows (Dahm, 1992). Another nonintrusive have used a single video camera to record both the
technique under developmcm by blfles (1992) is the reference and signal images for each velocity
RELIEF technique, which aLsoappears to be limited component', this split-image technique reduces
to two velocity components m a plane, sim_h_ to resohition by a factor of two, but also reduces system
P1"V. cost and comple_xity. The data by McKenzie for
single channel point measurements on a rotating
A fourth concept for acquiring non-intnlsive wheel (1995) display the best absolute accuracy (on
rea,l-mne velocity measurements in a planar region the order of + 1-2 m/see) of any measurements m
called Doppler global velocimetry (DGV) has been date. McKenzie's more recent (1997) planar
patented by Komine (1990). This technique uses a imaging results of the velocity of the same rotating
pair of video cameras and an iodine vapor cell wheel displayed a lower level of accuracy (± 2-5
absorption line filter (ALF) for each velocity m/see).
component, to measure the average Doppler
frequency shift, averaged over each pixel, ofthe light Thus, it is clear that in a very short time
scattered off minute seed particles in a flow as they (approximately six years), DGV has developed to a
pass through aplanar sheet of laser light. This new point where capability has been demonsuated for
DGV velocity measurement technique is currently making non-intnmve mean flow velocity vector
being developed into an accurate instnm_nt by a measurements in a plane, in a variety of complex
group at the NASA Langley Research Center siaglephase flow fields of practical significance.
(Meyers et a1.,1991). This same group has also While current scanned DGV systems lackthe
funded work at Northrop (M_ers and Komine, accuracy or resolution ofconventional LV systems or
1991). prv (to date, documented as on the order of from 5
to 8 %, versus 1% for LV, at about 100 fffsec.),
Others are also developing related DGV DGV has proven in a very short time to be an
concepts; for example, Miles ¢t al (1991) have extremely flexible whole-field velocimetry
developed a filtered Rayleigh scattering (FRS) technique.
technique which shows potential for providing non-
intrusive velocity measurements without requiring Following the nomenclature of McKenzie
any seeding. An optically thick ALF is used to filter (1995), the basic equation relating the Doppler shift
frequenctyo the resolved velocity component is photodetectors. Also, improvements in accuracy
given by 5v = ( o_- i ).V/_., where 8v is the Doppler have been obtained by carefully optimizing and
frequency, V is the velocity vector, k isthe incident matching amplifier gains for each pair of
laser frequency, and o_and i are the observer and photodiodes, as well as by enclosing each PDV
laser propagation directions, respectively (see Fig. channel to reduce background scattered light
1). Thus, the resolved velocity component is in the intensities.
direction ofthe sum of oand (-i).
Achievingadequatetemperature stability ofthe
APPARATUS AND PROCEDURE sidearms oftheiodinecellshasbeenfoundby
researcherastNASA Langleytobe an essential
The present point PDV system closely follows requixemenftoraccuratoeperatioonfaDGV system.
the basic DGV configuration using two inch A temperaturceontrolsystemwhichissimilarto
diameteriodinecellwshichwasoriginaldleyveloped thoseusedbyNASA Langleyhasbeenimplemented
by Meyers et al. (1991), except that photodiodes are inthepresemsystem,whichiscomprisedofapairof
cun'endybeingused,alongwithfrontlensesand electrical band heaters which heat a lmllow bushing
pinholest,ocollecstcattereldightfromasinglepoint made from oxygen-fzee copper, which mounds the
in a seededflowwhich isilluminatebdy a CW iodine cell except for the two optical windows and
Argon ion laser.Laserfrequencyhas notbeen thesidearm. The sidearm hasbeenthermally
activelcyontrollebdu,tinsteaadreferencieodinecell "grounded"byacopperwirewhichisbondedtothe
hasbeenusedtocompensateforanychanges(dueto tipofthesidearm,and thenboltedtotheoptical
laserfzequencydrifti)nthevoltageratiosforthe breadboardonwhichtheDGV systemismounted.
iodine cellswhich view the flow and receive the Theentiresystemhasbeenenclosedinan insulated
Doppler shifted scattered light A laser spectrum box(asshowninexplodedviewinFig.3)toshield
analyzer has been used to monitor laser mode shape thecellfrom aircurrentsor room temperature
and detect the occurrence of mode hops. After a variations, and the optical windows of the ceil have
suitable warm-up period for the laser and spectrum been insulated f_romthe room air by phenolic tubes
analyzer (typically on the order of one hour, to fitted with additionalAR-coated crown #ass
achieveoptimumfrequencsytabilitfyr)e,quency windows,whichprotrudethroughthesidesofthe
on the order of 50 MHz has been observed over a aluminumbox.Thiscreateasheateddeadairspace
time period on the order of 30 minutes; this is close nexttotheoutsidoefthecellbytheopticawlindows
tothe resolutioonfthespectrumanalyzerandthe andpreventtsheformatioonfsolidiodinecrystalosn
claimedfrequencsytabiliftoyrthelaser. the optical windows. The Omega PID temperature
contzoller has been adjusted to achieve very stable
This reference iodine cell system and the Argon operation, where the copper sheathsurroundintghe
ion laser are shown in Fig. 2, along with the laser iodinecelltypicalloyperateastatemperaturwehich
spectrum analyzer and Argon ion laser. Neutral isnominally10°Cabovethesidearm temperature.
density filters and a beam expander are used to Thisensuresthatallsolidphaseiodinecollectisn
ensurethattheiodinecellisnotsaturatebdy the thesidearmofthecell.Cellshavebeenoperatedat
reference beam. A quarter wave plate hasbeen used stem temperatureosf 45°C,sinceabsorptiownell
slopeisamaximum there(McKenzie,1995).
tocircularly polarize the laser light which is used to
illuminate the flow. The layout of one of the two
FDV channels is shown in Fig. 3, while a schematic DataacquisitiosnoRwarehasbeendeveloped
inVisualBasic4.0toallowcontinuousmonitoring
of how the entire system has been configured for the
wheel velocity m_ts may befound inFig. 4. anddataacquisitioofnthecelltemperaturefso,rthe
The two PDV channelsinclude pinholes behind reference cell and each of the cells used in the two
each fxontlens,as implementedby Roehleand PDV channels. Long term drift in iodine cell
Schodl (1994); these pinholes act as spatial filters to temperauae has been measured to be on the orderof
set the size of the sensing region in the light _heet, + 0.1°C (the specified set point accuracy of the
as wen as to reduce the effects of secondary temperatme controlleorn)c,ethecellhaswarmedup
to itssteadyoperaRugtemperature.Shortterm
scattering by limiting the depth of field. Piano-
convex lenses have been installed in front of the fluctuations have been measm_ whichareon the
photodiodes (see Fig. 3), to ensure that all scattered order of 0_03-0.04 °C;this approaches the resolution
ofour 16bitthermocouple A/D board.
light from the sensing region is imaged onto the
To determinethe accuracyof the two- developed, toautomate the data reduction process.
component PDV system, a rotating wheel apparatus
has been developed, consisting of a 12 inch Calibration of the iodine cells has been
diameter, anodized circular aluminum disk which accomplished using a continuous scan of the mode
has been painted white, and mounted on a variable structuroef the Argon ionlaser, by mechamcally
speed DC motor. This wheel can achieve tip altering the tilt of the etalon through about 10-20
velocities of approximately + 28 m/see. A second mode hops, over a 20-30 second period. A typical
wheel, consisting Ofa 6 inch diameter lexan disk time history of the voltage ratios for the reference
mounted to a Dremel tool can achieve tip speeds of and the two signal channels for this process isshown
about I00 m/sec, but this apparatus has not been in Fig. 6. It is generally noted that the signal-to-
used inthe present studydue to excessive wobble. In reference voltage ratio for each iodine cell varies
addition, a calibration procedure similar to that continuously between mode hops. Occurrence of
which has been used by NASA Langley personnel, mode hops has been detected by a sudden jump in
where the laser is mechanically mode hopped by reference photodiode voltage. The ratio valuefor
tilting the etalon has been utilized to calibrate the any one mode hop may be computed as an average
iodine cells. value, the value at the left end of that mode, or the
value at the right end of the mode; the best results
A 1.5 inch diameter, fully-developed turbulent have been obtained using the average value (James,
pipe flow apparatus has been developed, as has a 1997). It has been found that this continuous scan
small grid turbulence flow facility. Also, a jet mode hop calibration technique offers better
facility is available which has interchangeable accuracy than an earlier technique, where individual
convergent nozzles with exit diameters of 0.375, 0.5, ratio values were measured after each mode hop of
and 1.0 inches, and which can attain exit velocities the laser. This is because the cell temperatares
up to 120 m/sec, with very low exit turbulence levels. cannot change significantly over the 20-30 second
A conventional LV data set for this jet, as well as for time period required to perform a scan. Also, the
a companion annular jet, has been given by effects due to variability of where one stops the
Kuhlman (1994). Flow seeding has been achieved mechanical screw adjust on the etalon tilt screw are
using a commercial fog machine m the present work. minimized by this technique. Significant further
A second technique using a soldering iron to improvement incalibratiaocncuracy has also been
vaporize the fog fluid is also available (McKenzie, obtained by averaging several of these individual
1995). continuous scan mode hop calibrations together.
This improved calibration consists of several (from 3
A computer-controlled, three-axis traversing to 6) continuous mode hop calibration dam sets for
system has been developed (Fig. 5), as described in each cell. A single cell calibration data RIe isformed
the thesis byRamanath (1997), for use in positioning by "sliding" all mode hop calibrations for any one
the flow facilities with respect to the fixed, two cell,to overlay them on one a.d_itranly-seleaed
channel PDV system, so that velocity contours may calibrationof the seL This procedureis
be mapped out in a plane or volume. This traverse accomplished by linear interpolation, and is
allows movement in a volume which is two feet by a necessary because of laser drift between mode hop
foot and a half in a horizontal plane, by one foot in calibrationwsh,ere the ratiovaluefor the n_ mode
the vertical direction. Accuracy of a single traverse hop for anyone cellwillchange, especially when the
move has been found to be better than 0.001" for room temperatuz¢variessignificantly.
typical moves on the order of a few inches
(Ramanath, 1997). Eight bit Hitachi CCD cameras and a Matrox
Genesis frame grabber are being used for the two
An 8 channel, 16 bit, simultaneous-sample- component DGV system, which is currently under
and-hold IOTech A/D board is used for digital data development.Sincebotha2-componentpoimPDV
acquisition of the photodetector output voltages for systemas wellas a 2-componentimagingDGV
the reference iodine ceil, and for the two PDV systemwill eventually be available,itwill be
channels. The RMS noise level for this board is + possible to make direct comparisons betweenthe
0.3 mV on a I0 volt scale. Windows-based data PDV measurements and the scanned DGV
acquisition soRwax¢ has been developed (again, measurements in the same flow fields, measured
using Visual Basic) for this boar& In addition, from the same viewing directions, with comparable
companion VB data reduction programs have been smoke particle size and laser illumination levels.
This will provide direct experimental documentation velocitdyatasetswereacquireda,llusinga single
of the amount of additional error introduced by the cellcalibratidoantarun(James,1997).A typical
CCD video cameras. exampleoftheresultindgataisshown inFig.7,
wherethemeasuredPDV velocitmyagnitudesfor
RESULTS channelsI and 2 areshown plottedagainstthe
known wheelomega-r.ObservedImearityisquite
Early data repeatability, as documented in the good.The standarddeviationosftheslopesofthe
thesis byRamanath (1997), was poor. The standard Linearcurvefitequationsfromthecorrectslopeof
deviation of the slopes of plots of the measu_ PDV exactly 1.0 were calculated forseveral different
velocity versus the known velocity of a rotating curve fit options for each of the seven data sets.
wheel was on the order of 8-15 %, even though the These errors ranged from a maximum of about*15
linearity of each individual data set was quite good %fora linear curve fitof the cell calibrations (onthe
(on the order of + 1-2 m/_c, out of 58 m/sec). same order of error as for the earlier data), to a low
Similar resatlts were initialloybtainedby Iames of ±1-2 %for a fourth-order curve fit. Specifically,
(1997). However, the improved cell calibration from the individual slope results listed in Table 1,
procedures described above have significantly the deviations of the measured sensitivities were 1.4
increased the accuracy of the present PDV system. % for channel 2 and 2.3 % for channel 1 using
fourth- order curve fits to the calibration data. The
Typical examples of the present results for the accuracy to which the viewing angle could be
rotaRng wheel will now be presented; these damare measured was less for the channel which viewed the
presented in much more detail in the thesibsyJames wheel off of the mirror (channel 1); this is believed
(1997). Dam has been acquired in two different to be the explanation for the larger error in
fashions: Rrst, the 2-component PDV system was sensitivity or slope forthischannel.The actual
configured so that we could acquire two standard deviations of the slope values are somewhat
simultaneo_, but inde_ndent, measurements of the smaller, indicatintghatthereissome biaserrorm
wheel velocitmyagnitude, from two slightly different theseresultsc;hannel2 slopeshave a standard
viewing directions. To dothist,hesystemwassotup deviation of 1.1%, while channel 1slopes have a
so thatboth PDV channels had relativelgyood standard devia_on of 1.5 o/_ Since the total range of
sensitivities m the wheelvelocity directionO.ne wheel velocity for these measurements is about 57
channel (channel 2) was set up with a viewing m/sec, these oi_erved 1-2 % errors correspondto
directiownhich was at an angleof approximately 42 velocity errormagnitudes of approximately e0.6-1.2
degrees from the laser propagation direction. The m/sec, which isquite good. Also, the standard
other channel (channel 1) viewed the wheel at an deviations of the actual PDV velocity data points
angle of approximately 75 degrees, by imaging the fromtheleastsquareslinearcurvefitshavebeen
wheel off of a mirror which was mounted on the computed,aslisteidnTableI,and theseerrorsare
breadboard which held the channel 2 oFdcal evensmallerthantheslopeerrors.Forchannel2,
components (see Fig. 4). The wheel was inclined the data for all sevenrunsdisplaya standard
slightly (about 5 degrees) to the laser propagation deviation from a linear fit of 0.7 % (i-0.4 m/soc),
direction. In this configuration, the re_ting while channel I displays a standard deviation from a
velocity datahave been convcxted towlu_ velocities lineafritof 0.5 % 0:0.3 m/sec). Again,thislevel of
byassumingthatthedirectioonfthewheelvelocity linearity is quite good. Sincetheerrorsintheslopes
was known. In thesecondconfigurmiont,hetwo of the meamred velocity versus omega-r were
PDV channels were set up with widely differing mmllest using the fotmh-order polynomial curve fits
viewing angle, (al_mately 126 degrees from the to the average ratiodata,allmb*_luem datahas
laser propagation direction for channel 1, and beenreduced using this method. Ithas been observed
approximtely 42 degrees for channel2), and the that the reference cell is not able to consistently
two PDV velocity measurements were used to resolve determine the zero velocity, thus, for the present
orthogonalx- and y-velocitcyomponents,from results zero velocity has been fixed by a
which the wheel velocity magnitude was computed measmement of all voltage ratio values just prior to
asthe square root of the sum of the squares of the and after the actual data acquisi_on- Other
components. researchers using scanned systems have had similar
problems, which they have addressed by imaging a
With the 2-component PDV system set up in zerovelocityregionsomewhere in each camera
the first configurationa,series of seven wheel image (McKenzie, 1996, and Reinath,1996,
personacol mmunication).
speed has been set has been checked by using a
strobe and measuring the time between voltage
With the 2-component PDV system set up in spikes in the output of one of our PDV signal
the second configuration, a series of twelve wheel photodiodes, using an oscilloscope. The reading
velocity data runs have been acquired, using five error for these time measurements was estimated to
different cell calibrations (James, 1997). An be no greater than 4.0.5 %, which isabou_ the same
example of this data is shown in Fig. 8, while the as the observed accuracy to which the individual
individual orthogonal velocity measurements axe PDV data points fit to a least-squares straight line.
shown for this run in Fig. 9. The individual slopes To do anybetter thanthis at checking the linearity of
and standard deviations of the data f_m a linear fit response of our instrument, the individual wheel
for each runhave been given in Table 2. Here the speed settings would need to be measured with
standard deviation of the slopes of the linear curve greater accuracy, for each datapoint.
fits to the data is 1%, and the data points exhibit
deviations from the linear curve fits with a standard 2. The level of zero velocity drift of our 2-
deviation of 1.1% (4.0.65 m/sec). Again, this is felt component PDV system has been measured for a
to be quite good accuracy. The accuracy of the period of 30 minutes, as shown in the Fig. 10. The
channel 1data, which is less sensitive to the wheel observed drift in zero velocity ison the orderof_+1.5
velocity, is not as good as that of channel 2. As a m/sec;this is thought to be primarily due to drift in
result, the accuracy of the computed x-velocity cell stem temperatures, as shown in Fig. Il, where
component, normal to the laser propagation the diffcren_ between the stem temperature for the
direction, is not as good as the accuracy of the PDV signal channel, and the reference cell stem
computed y-velocity component (Fig. 9). The correct temperature, is shown foreach channel, for the same
sensitivities to the x- and y-velocity components, as time period as the zero velocity dam run(Fig. 10).
shown in Fig. 9, were co_5 °) and -sin(5°): The similarity of the shapes of these two graphs
indicates that this error is due primarily to the
During efforts to improve the accuracy of the combination of the accuracy of the ceLl stem
point DGV system, typical RMS fluctuation levels of tempemtme controllers and the accuracy of the
the voltage signals firom the photodiodes have been cah_orafioncurves. Itisalso felt thatthis_ in cell
monitored, along with the RMS fluctuation levels of stem temperature difference is a significant
the computed ra_o of signal-to-reference voltages. contributorto theobservederror m sensitivity of the
For the reference system, RMS voltage fluctuations two PDV channels. However, no correlation has
typically are on the order of 0.5 % of the mean been found between the slope data given in Tables 1
voltage for each photodiode, butthe RMS fluctuation and 2, andthe stem temperature difference values for
in the ratioisapproximately 0.2 % ofthe mean ratio each data run. It is noted that the present zero
value. Similar percentage fluctuations in the ratio velocity _ (+ 1.5 m/sec) is reduced relative to the
value have been observed during experiments using earlier results presented byRamanath (1997), of _.4
the rotating wheel. Recently, a _x_ple math model m/_c; this is due to the improved accuracy of the
of thisphenomenon hasbeen proposed by Ramanath, present cell calibra_ons.
where the individual signals fi'om the photodiodes
are modeled as sine fimction_ each of which can 3. Using four of the five calibration curves
have offset amt/or phase errors. The ratio of these which have been used to reduce the second series of
two model signals can show increased, "spiky" wheel velocity data, the repeatability of the
fluctuation levels, as is sometimes observed in the calibration curves from day-to-day has been
datawhen the raw voltage levels are small. Offset investigated (James, 1997), by forcing pairs of
errors may occur due to _es or changesin corresponding calibration curves to overlay exactly
the detector dark or background voltages; such at the middle of the ratio rangewhich was actually
errors become more significant as the signal level used to reduce the wheel velocity data, and
decreases. computingthe differencien computed velocities
which would result at the top and bottom of these
An analysis of the major error sources for the ratiorangesff the two different cellcalibrations were
present remits has been performed, as very briefly used to rednce the same wheel velocity dataset. (See
summarized below. Fig. 12 for an example.) The resulting errors were
found to be between 2.3 and 0.3 m/sec, with and
1. The accuracy to which the rotating wheel average error of 0.7 m/sec. (4. 1.2 %). These errors
r 6
dueto using different cell calibrations are on the Accuracy of the present PDV system, based on
same order as the observed error in slopes of the the rotating wheel velocity results, has been
wheel velocity data remits. As a result, itisbelieved documented tobe on the order of -. 0.6 m/sec over a
that while the current cell calibration procedure has velocity range of 57 m/sec (ie, approximately + 1%
led to a significant improvement m our results, the of full scale). Linearity of the present PDV system,
primary error source m the current PDV system is again basedon the wheel data, has been documented
still the ac_'xLracyof the cell calibrations. To improve to be on the order of ± 0.3 m/sec, which is on the
_s accuracy farther, work is underway to order of the accuracy of the individual wheel velocity
implement the curve-fitting procedure described by settings; this lmearity ison the order of 0.5 % of the
McKenzie (1995), which uses the theoretical iodine measured velocity range. Both of these observed
cell absorption curve calculated by the Forkey accuracy measurements are considerably better than
theoretical cell absorption model to curve fit the those documented todate by other researchers.
actual cell calibration data. It is felt that this
procedure will remove the judgment which is Preliminary two-component PDV velocity data
required in the current polynomial curve fit have been presented for a fully-developed turbulent
technique, where the user must decide which po_on pipe flow, at a Reynolds number of approx_11ately
of the cell calibration dam to include m the curve fit 76,000. Turbulence intensity values agree well with
process. Also, from Fig. 12itis notedthat the useful earlier hot wire data, and mean axial velocity data
range ofthe developed PDV system isonthe or(_r of agree reasonably well with pilot tube results.
I00 m/sec; this range could be increased by the However, radial velocity results show a consistent
addition of a neutral buffer gas to the iodine vapor offset error which is on the order of ten percent of
(Elliott, et al., 1994). A reduction in this useful the mean axial velocity; the reasons for this errorare
range, in hopes of increasing the resolution of the at present uncertain, btrt are beheved to be due to
Doppler velocimeter, may be possible through use of inaccuracy inthe determination of zerovelocity.
a Cesium Faradayfilter (Bloom, et at., 1993).
It isplanned that the developedtwo-compenent
An example of preliminary two=compotmnt PDV _tation will be utdized to obtain
PDV data obtaLrmdfrom a traverse across the exit of further velocity datain the fully-developed turbulent
the fully-developed pipe flow apparatus, ata nominal pipe flow and axisymmetric jet flow facilities. These
Reynolds number of 76,000, is shown in Figs. 13 data _ be compared with conventional LV data.
and 14. These data have been obtained at a Sim/hr studies are also planned using the two-
sampling rateof 5kHz, with a data record length of componentscannedDGV systemwhichiscurrently
10(points (2 seconds) at each measurement location. underdevelopmentT.heultimatgeoalofthepresent
The axial mean velocities agree reasonably wall researcihstooptimizethesystemaccuraciefsorboth
withresults fi'omapilot tube survey, butradial mean the point PDV and scanned DGV systems, and then
velocities currently display an offset error on the to carefully documem the accuracies of both
order of 2-4 m/sec (Fig. 13). Turbulent velocities optimizedsystems,andquantifythedominant error
(Fig. 14) agree well with the hot wire dataof Lm_ sources foreach.
(1954). There arc significant difficulties near the
pipe wal/_ _ due toreduced signal-to-noise levels ACKNOWLEDGMENTS
due to less smoke, as well as to reflections of the
scattered light off of the pipe wails. However, at The presentwork is being supported under
present the greatest di_culty appears tobeobtmnm"g AFOSR/DEPSCoR Grant F49620-94-1-0434, Dr.
accurate, reliable zero velocity ratio values. James M. McMichael, technical monitor. Senthil
Ramanath received partial support under NASA
CONCLUSIONS grant NAGW-4464. The authors are grateful for
the technical assistance of YimMeyers, Joe Lee, Rich
A two-component Point Doppler Velocimeter Schwartz, Angelo Cavone, and Gary Fleming at
(PDV) system has been developed, aad its accuracy NASA Langley _ Center, and Bob Mc_Ke_e
hasbeen investigated by measuring the velocity of a andMike Reinath atNASA Ames Research Center.
rotating wheel. Also, prelimina_ two-component
PDV velocity data have been presented for a fully- REFERENCES
developed tu_oulent pipe flow.
Adrian, R.J. and Yao, C. S., "Development of
PulsedLaser V¢locimetry for Measuremem of Mcyers, J.F. and Komine, H., "Doppler Global
Turbulem Flow," Proc. of the Eighth Biennial Velocimctry: A New Way to Look at Velocity," 4th
Symposium on Turbulence, Univ. of Missouri, RolIa, Imemational Conference on Laser Anemometry,
1983,pp.170-186. Advances and Applications, Aug. 5-9, 1991,
Bloom,S.H.,SeaxcyP,.A.,Choi,K.,K.remer, Cleveland,OH, pp. 289-296 of Conference
IL,andKorevaar,E.,"HelicopteFrlumeDetection Proceedings.
by Usingan Ultranarrow-BanNdoncoherentLaser Meyers, J.F., Lee, J.W., and Cavone, A. A.,
DopplerVelocimeter,O"pticsLettersV,ol.18,No. "Signal Processing Schemes for Doppler Global
3,Feb.1993,pp.244-46. Velocimetry," 142 International Congress on
CIancy,P. S. and Sam/my, M., "Multiple- Instrumentation in Aerospace SimulatioFnacilities,
ComponentVelocimetriynHighSpeedFlowsUsing RockvilleM,D, Oct.27-31,1991.
PlanarDopplerVelocimetry,p"aperAIAA-97-0497, Miles, R.B., "RELIEF Measurements of
AIAA 35_ AerospaceSciencesMeeting,Jan.6-10, Turbulence," AFOSR/ONR Grantee and Contractors
1997,Reno,NV. Meetingon Tuflmlence_h, Chicago,IL,June
Dahm, W.J.A.,"ScalarImagingVelocimetry 10-12,1992;pp.62-64ofWorkshopProceedings.
StudieosfTurbulenFtlowStructuraendDynamics," Miles, R.B., Lamport'W.R., and Forkey, J.,
AFOSR/ONR Grantee and Contractors Meeting on "Instantaneous Velocity Fields and Background
TurbulenceResearch,Chicago,IT.,June 10-12, SuppressiobnyFiltereRdayleigh Scattering," paper
1992;pp.73-76ofWorkshopProceedings. AJ.AA-91.0357,AIAA 29th Aerospace Sciences
ElliotG,.S.,Samimy, M.,and Arnette,S.A_, Meeting, Jan. 7-I0,1991, Reno, NV.
"Details of a Molecular Filter-Based Velocimetry Morrison, G. L., Gaharan, C.A., and DeOtte,
Technique," paper AIAA - 94-0490, AIAA 32nd IL E., Jr., "Doppler Global Velocimetry: Problems
Aerospace Sciences Meeting, Jan. 10-13, 1994, and PitfallsS,y"mposium on LaserAnemometry:
Reno, NV. AdvancesandApplicationJsu,ne19-23,1994,Lake
Forkey, LN., Lempert, WR., Bogdonoff, S.M., Tahoe, NV, andpublishodinASME FED Vol.19I,
Miles, R.B., and Russell, G., "Volum_ic Imaging pp.I-8.
of Supetsomc Boundary Layers Using Filtered Ramanath, S., "Development of a Point
Rayleigh Scattering Background Suppression," paper Doppler GlobalVelocimeter (DGV)," MS Thesis,
AIAA-94-0491,1994, Reno, NV. West Virginia University, MAE Department, 1997.
Hoffenberg, R. and Sullivan, I. P., "Filtered Roehle, I. and Schodl,R.,"Evaluation ofthe
Panicle Scattenng: Laser VelocimeU7 Using an Accuracy of the Doppler Global Technique," paper
Iodine Filter," AS/viE FluidsEngineering Division C485/046, Seminaron OpticalMethodsandDam
Summer Meeting, Washington, DC, June 20-24, Proces_ng in Heat and Fluid Flow, Apr. 14-15,
1993. 1994, City Univ., London, UK.
Jam_, K., _eterminadon of the Accuracyofa Smith, M.W. and Northam, G. B., "Application
Two-Componem Point Dopplex Vdochnctry of AbsorptionFilter-Planar Doppler Velocimctryto
System," MS Thesis, West Virginia University, SonicandSupersonicJets," paper AIAA-95.0299,
MAE Department' 1997. AIAA 33rd Aerospace Sciences Meeting,Jan. 9-12,
Komme, I-L,U.S. Patent No. 4, 919, 536 Apr. 1995, Reno, NV.
24, 1990.
Kuhlman, J.M., "Turbulence Measuremems in
Annular Jets Using Laser Velocimetry", ASME
Syrup. Laser Anemomeuy: Advances and Table I Slopedataforr_aing wheel,setup with
Applications, Lake Tahoe, NM, June 19-23, 1994. mirrorandusingknownvelocidtiyrection
Laufer, J., "The Structure of Tm'bulence in
Fully DevelopedPipeFlow,"NACA TR 1174, 1954. Run SIo_ DoAafionfi'olmineafrit
McKcnzie,ILL.,"MeasurementCapabilitioefs (ChI,Ch 2) (Ch1,Ch2,inm/sec)
Planar Doppler VelocimeW/Using Pulsed Lasers,"
paper AIAA-95.0297, AIAA 33rd Aerospace I 1.00731,.0270 0.190,.26
2 0.97750,.9916 0.220,.35
Sciences Meeting, Jan.9-12, 1995, Reno, NV.
3 0.98761,.0090 0.290,.40
McKenzie, R. L., "Planar Doppler Velocimew/
4 0.986!0,.9988 0.330,.50
Performance in Low-Speed Flows,"paper AIAA-97- 5 0.96031,.0046 0.370,.45
0498, AIAA 35_ Aerospace Sciences Meeting, Jan. 6 0.97471,.0019 0.240,.46
6-10, 1997, Reno, NV. 7 0.99111,.0142 0.35,0.26
8
Table 2 Slope data for rotating wheel; two-channel PDV
setup
_ko s--,.:/
Run Slope Deviation from
i 0.9808 linem0"f.i6t7(zlqsec) d_,_=, ___X.?/O_.x_ ==_e=='
.........
2 0.9981 0.50 v -
3 0.9941 0.63 _,J a j
5 1.0013 0.45 n,_ V
6 1.0080 0.99 _ --
7 1.0214 014
48 01..09097936 01..3345 [ z,_., _m
9 0.9961 1.03
I0 0.9928 0.42
11 0.9904 0.50 Fig, 1 Determination of Point Doppler Vel_
12 0.9989 0.69 (PDV) Doppler _ frequen_
$ignGl
Iodine Ceil PhotodeteC%or
Be_a_ Exl_a_er \ Lees / - /e_DerQ_ure Control Unit
(Removed to Expos
IlnodSlutQr_tl Cneglt)Box L_see
Dens| %y FIt%er
HGIirmroloral Ie_n_4our_% %rGl
,,ha Gl_mtted _,,,_jIIl I [ __" ///
NeutrQt _ensi %y __ql_-_-_PY rex _ass
Fiit_
Pyrex c,tass _¢_-ter-eave Prate
Fig.2 Apparatm for refcren_ iodil_ edl _ laser, a_i _ctram a_alyz_
SIg_t
_ / _ _--_J,_q_l_mW\ // I I [ (Removed to Expose
Fig. 3 _forot_PDVcham_
9