Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Pressure Power Spectra Beneath a

Supersonic Turbulent Boundary Layer


Steven J. Beresh,* John F. Henfling,†† Russell W. Spillers,‡‡, and Brian O. M. Pruett§
Sandia National Laboratories, Albuquerque, NM, 87185

Wind tunnel experiments up to Mach 3 have provided fluctuating wall-pressure spectra


beneath a supersonic turbulent boundary layer to frequencies reaching 400 kHz by
combining signals from piezoresistive silicon pressure transducers effective at low- and mid-
range frequencies and piezoelectric quartz sensors to detect high frequency events. Data
were corrected for spatial attenuation at high frequencies and for wind-tunnel noise and
vibration at low frequencies. The resulting power spectra revealed the Ȧ-1 dependence for
fluctuations within the logarithmic region of the boundary layer, but are essentially flat at
low frequency and do not exhibit the theorized Ȧ2 dependence. Variations in the Reynolds
number or streamwise measurement location collapse to a single curve for each Mach
number when normalized by outer flow variables. Normalization by inner flow variables is
successful for the Ȧ-1 region but less so for lower frequencies. A comparison of the pressure
fluctuation intensities with fifty years of historical data shows their reported magnitude
chiefly is a function of the frequency response of the sensors. The present corrected data
yield results in excess of the bulk of the historical data, but uncorrected data are consistent
with lower magnitudes. These trends suggest that much of the historical compressible
database may be biased low, leading to the failure of several semi-empirical predictive
models to accurately represent the power spectra acquired during the present experiments.

Introduction
The accurate measurement of the fluctuating wall-pressure spectrum beneath a supersonic turbulent boundary
layer has proven elusive for many decades. Dolling and Dussauge [1] discuss many of the reasons why this is the
case. The principal problem is sensor frequency response; the temporal frequencies of interest in high-speed flows
typically extend beyond the capability of most pressure sensors. Moreover, resonance of both the sensor diaphragm
and the cavity often occurs at frequencies that contaminate the measurement. Spatial resolution of the sensors also
poses a problem, as the spatial scales of the flow in which significant energy is contained are commonly much
smaller than the sensor size, leading to a low-pass filtering of the signal. Electronic noise also is a routine difficulty
in wind tunnel environments, hindering the detection of subtle effects. The combination of these difficulties often
limits measurement frequencies to those below about 100 kHz, and sometimes considerably less.
Despite these challenges, numerous efforts have been made to acquire unsteady wall pressure data for
supersonic boundary layers, on a variety of geometric surfaces. Many such experiments have been cited by Dolling
and Dussauge [1], Dolling and Narlo [2], and Laganelli and Hinrichsen [3], but the results they report exhibit an
alarming degree of scatter for nominally compatible measurements. Partly, differences may arise because the
experimental conditions themselves are dissimilar or insufficiently reported, or are influenced by unknown effects.
But some of the data scatter certainly results from the variety of measurement errors discussed earlier. As a result,
universal trends in the data may be obscured, and the development of engineering models and computational

*
Principal Member of the Technical Staff, Engineering Sciences Center, Associate Fellow AIAA, correspondence to: P.O. Box
5800, Mailstop 0825, (505) 844-4618, email: sjberes@sandia.gov
††
Distinguished Technologist, Member AIAA
‡‡
Senior Technologist
§
Senior Technologist

This work is supported by Sandia National Laboratories and the United States Department of Energy. Sandia is a
multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of
Energy’’s National Nuclear Security Administration under Contract DE-AC04-94AL85000.

-1-
American Institute of Aeronautics and Astronautics
simulations may be hampered by an incomplete understanding of the behavior of supersonic wall pressure
fluctuations.
Experiments have been conducted in Sandia’’s Trisonic Wind Tunnel up to Mach 3 to provide fluctuating wall
pressure data under well-characterized flow conditions to frequencies exceeding 100 kHz. Carefully acquired, these
data can play an important role in reconciling the conflicts present in the historical data. An earlier document [4]
has detailed improvements in the pressure sensor capability by developing an innovative technique to merge
measurements from three sensors. Data have been acquired using piezoresistive silicon pressure transducers
effective at low- and mid-range frequencies, then supplemented by piezoelectric quartz sensors capable of detecting
very high frequency events. The two sensor types were dynamically calibrated against a condenser microphone
reference standard, then combined into a single curve describing the wall pressure spectra. The effects of wind
tunnel noise and vibration at low frequencies were reduced using an adaptive filter technique similar to Naguib et al
[5] and the effects of spatial attenuation at high frequencies were compensated using the Corcos correction
technique [6]. Using this approach, measurements were shown to be accurate to 400 kHz and to produce results
consistent with more reliable incompressible experiments, but at the same time they suggested that the historical
database of supersonic pressure fluctuation intensities is biased low.
The present paper utilizes this improved measurement capability to acquire pressure power spectra under a
variety of flowfield conditions, then compares the results with the historical database and common engineering
models. Though even unsteady pressure measurement often is regarded as routine in wind tunnel experimentation,
unrecognized error sources have repeatedly contaminated past experiments, and thus careful conduct of the present
study can yield results that will clarify murky trends in the historical literature.

Historical Database
The intensity of the pressure fluctuations is commonly characterized by the r.m.s. of the unsteady signal, ıpƍ,
and normalized by either the edge (or freestream) dynamic pressure qe or the wall shear stress IJw. High-speed
experiments generally have preferred normalization to qe because of its ease of measurement in comparison to IJw
and its practicality in engineering applications. The theoretical justification for normalization to IJw is stronger, but
the measurement uncertainty (or predictive inaccuracy) usually is considerably greater, and furthermore a
dependence upon Reynolds number is known.
Even the measurement of incompressible fluctuating pressure intensities on a smooth flat plate has posed a
formidable challenge. Numerous efforts to compile a database of high-speed pressure fluctuations have cited an
incompressible value of ıpƍ/qe=0.006 [e.g., 7-9], a value apparently originally attributable to the work of Willmarth
[10, 11] and Bull [12] and Shattuck [13]. Although it was widely accepted for a long time, evidence accumulated to
suggest that this value was biased low due to insufficient sensor spatial resolution [6, 14, 15]; see reviews by
Willmarth [16] and Bull [17]. Subsequent experiments gradually raised the pressure intensity value. Bull and
Thomas’’s study [14] suggests a correct value near ıpƍ/qe=0.008 but may suffer from perturbations induced by a
pinhole microphone [17]. Schewe [15] shows that a sufficiently small microphone will return ıpƍ/qe=0.010, a result
supported by Lueptow [18]. Gravante et al [19] find ıpƍ/qe varies from 0.008 to 0.010 as Reș decreases, and Goody
et al [20] find about 0.009. Farabee and Casarella [21] integrate their measured power spectra and the estimated
spectra beyond their measurement capability to determine ıpƍ/qe=0.009, which, in the view of both Bull [17] and the
present author, probably is the most correct estimation of the incompressible value of ıpƍ/qe.
Supersonic and hypersonic measurements of ıpƍ/qe are considerably more challenging. The difficulties
encountered for incompressible measurements regarding spatial resolution and frequency response only become
more acute as freestream velocities rise and boundary layer thicknesses typically diminish due to the smaller size of
most testing facilities. Nonetheless, the importance of pressure fluctuations to flight vehicle performance has
motivated numerous experiments, which are collected in Fig. 1. Data are given in terms of the edge Mach number,
Me, as the compilation includes experiments conducted on cones and sting-mounted flat plates, but for
measurements on a wind tunnel wall, this is equivalent to the freestream Mach number, M’. The range of
incompressible values is based upon the discussion in the previous paragraph and is summarized by a single data
bar. Most data have been obtained on a wind tunnel wall to yield the thickest boundary layers, as in the present
experiment, but nominally compatible measurements from flat plates, cones, and ogive cylinders are included as
well. Given that the data from these latter geometries may be subject to additional influences or uncertainties such
as post-bow-shock conditions, transitional/turbulent flow state, thinner boundary layers, and leading edge effects,
these measurements have been displayed as hollow data points; measurements on wind tunnel walls are denoted by
solid data points. Some of the supersonic data are obtained from experiments nominally concerned with
shock/boundary-layer interactions, but also contain data in an undisturbed flowfield that may be included here. The

-2-
American Institute of Aeronautics and Astronautics
Fig. 1: Historical database of pressure fluctuation intensity as a function of Mach number, normalized by
dynamic pressure. Solid points show data measured on a wind tunnel wall; hollow points measured on a
flat plate, cone, or ogive cylinder. Four semi-empirical predictive equations are shown as well.

bulk of the measurements presented in Fig. 1 were obtained in the 1960’’s and 1970’’s and did not have available the
measurement technology now commonplace. Most of the more recent measurements are extracted from the
shock/boundary-layer database and are limited in Mach range.
Figure 1 dramatically illustrates the impressive scatter of several decades of nominally compatible
measurements. In fact, nearly the only conclusion that may be drawn with any confidence is that the pressure
fluctuation intensity diminishes with rising Mach number, and even then the strength of this trend is ambiguous. No
differences are evident between the data acquired on a wind tunnel wall and those measured on other surfaces.
Also shown in Fig. 1 are curves for several semi-empirical predictive equations for the pressure fluctuation
intensity. The simplest is Lowson’’s equation, which is solely a function of the Mach number [8]. Chaump et al [35]
later found that Lowson’’s prediction was consistent with the data for supersonic Mach numbers, but developed an
extension to his curve for Mach numbers exceeding 5, also based solely upon the Mach number. A more complex
predictive equation was developed by Laganelli et al [50] that incorporated effects due to the wall temperature. Two
curves are shown for the Laganelli prediction to account for this dependence, one for Tw=Taw and one for the
bounding condition of Tw«Taw. Wall temperature effects within the historical data are discussed subsequently.
Generally speaking, however, none of predictions well represents the experimental data, which hardly is surprising
given the broad scatter of the data points.
The scatter of the data in Fig. 1 may arise from trends not taken into account, such as those due to varying
Reynolds number. In incompressible flow, no evidence has been found of a relationship between ıpƍ/qe and the
Reynolds number based either on boundary layer thickness, Reį, or momentum thickness, Reș; however, some
variation may exist for small values of Reį or Reș but vanish once moderate Reynolds numbers are reached [19].
Conversely, a trend with Reynolds number may be present for high-speed flows. Raman convincingly finds that
ıpƍ/qe falls as a function of Reį*-0.2 [43, 44] when the Mach number is constant, where į* is the boundary layer
displacement thickness, which is quite close to Maestrello’’s assessment that ıpƍ/qe falls as Reį-1/4 [33]. Kistler and
Chen also find a decreasing ıpƍ/qe as Reį* falls at constant Mach number [32].
To ascertain whether a Reynolds number effect is present in the historical database and may account for the
scatter, the data points of Fig. 1 have been recast in Fig. 2 with their color a function of their Reynolds number.
Data points have been excluded for which the Reynolds number was not provided or could not be reasonably
estimated. Figure 2a shades the points based upon Reį and Fig. 2b based upon Reș. Though Reį is not as descriptive

-3-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 2: Data points from Fig. 1 shaded by the Reynolds number based upon: (a) 99%-velocity boundary layer
thickness; (b) boundary layer momentum thickness.

(a) (b)
Fig. 3: Data points from Fig. 1 shaded by characteristic sensor parameters; (a) non-dimensional cutoff frequency;
(b) cutoff frequency due to spatial resolution.

of the boundary layer state as Reș, į is more commonly measured and, arguably, also is easier to evaluate in a
compressible flow as compared to į* or ș. Trends as a function of Reį* generally track those as a function of Reș,
and hence are omitted for brevity. Unfortunately, any pattern is difficult to identify. It appears in Fig. 2a that some
lower values of ıpƍ/qe correspond to greater Reį, but this only is definitive for the Raman data points [43, 44].
Similarly, Fig. 2b appears to show a cluster of data points at lower ıpƍ/qe when Reș is larger, but again this is hardly
definitive.
Another possibility is that any trends found in Fig. 1 are not due to the physics, but rather, due to measurement
limitations. The frequency response of a pressure sensor can be described in non-dimensional form by gį/U’, where
g is the cutoff frequency characteristic of the sensor [1]. The spatial resolution of a sensor induces a cutoff
frequency as well due to attenuation of wavelengths smaller than the sensor’’s sensitive area, which can be
characterized by Uc/ʌd [6], where Uc is convection velocity near the wall and can be reasonably approximated as
Uc/U’=0.6 [1] and d is the sensor diameter. Figure 3 recasts the data of Fig. 1 based upon these two sensor
parameters, shown in Figs. 3a and 3b respectively, excepting those data points for which insufficient information
was available to determine either gį/U’ or Uc/ʌd. A calculation of values for gį/U’ is complicated by the fact that

-4-
American Institute of Aeronautics and Astronautics
Fig. 4: Historical database of pressure fluctuation intensity as a function of Mach number, normalized by
wall shear stress. Solid points show data measured on a wind tunnel wall; hollow points measured on a flat
plate, cone, or ogive cylinder.

in many cases the frequency response is an estimate


rather than a rigorously determined value, but
nevertheless a clear pattern emerges. Figure 3a does in
fact show an identifiable trend of larger values of ıpƍ/qe
arising from sensors with larger cutoff frequencies.
This is not surprising, as it stands to reason that sensors
capable of measuring a broader portion of the frequency
spectrum will be sensitive to more pressure fluctuations
and hence record a greater total intensity. It also is
worth recognizing that achieving a better frequency
response can be accomplished not only by fast-response
sensors, but alternatively by obtaining measurements in
a fairly thick boundary layer; this appears to have been
key to the apparent success of some of the earliest
experiments.
On the other hand, Fig. 3b does not indicate any
distinct trend of ıpƍ/qe as a function of spatial resolution, Fig. 5: Data points from Fig. 4 shaded by sensor non-
which suggests that sensor frequency response is the dimensional cutoff frequency.
dominant issue for high-speed flows. This observation
is initially surprising given the clear need in
incompressible flow for adequate spatial resolution [15, 18] or corrections for spatial attenuation [6]. However,
what is actually demonstrated is that improvements in spatial resolution are not useful without corresponding
improvements in sensor frequency response. For example, the lines of the Raman [43, 44] data points at Machs 5.2,
7.4, and 10.4 have large values of both gį/U’ and Uc/ʌd, and hence return the largest hypersonic values for ıpƍ/qe.
Conversely, the cluster of Harvey et al [26] data points between Mach 8 and Mach 9 shows a large value of Uc/ʌd,
but without a similarly large value of gį/U’ to match, lower values of ıpƍ/qe are returned. In a small number of
cases, the effective high-frequency limit due to the spatial resolution is less than the high-frequency limit of the

-5-
American Institute of Aeronautics and Astronautics
sensor itself; inserting the lower values due to spatial attenuation into Fig. 3a does not appreciably change the
results.
Some of the data scatter in Fig. 1 may be attributed to wall temperature differences between the experiments, as
suggested by the range of results predicted by Laganelli et al [50] as Tw/Taw varies. In practice, however, for most
moderate values of Mach number, Tw/Taw is fairly close to unity, and therefore the wall temperature is not a
satisfying explanation for those data points exhibiting large values of ıpƍ/qe at Me<5. The Tw«Taw Laganelli curve
really is meaningful only for hypersonic Mach numbers, at which prediction of ıpƍ/qe grows more difficult because
wall temperature ratios deviate more markedly from the adiabatic assumption. The Raman experiments [43, 44] had
a value of Tw/Taw=0.3-0.45, which could explain their generally elevated values. Conversely, Martelucci et al’’s
Mach 6.7 and 8 data points [34, 35] were acquired at Tw/Taw=0.48 and 0.3, respectively, but reside near the Laganelli
Tw=Taw curve, and similarly, the Chung and Lu [22] data point corresponds to Tw/Taw=0.39. The other hypersonic
experiments used indeterminate values of Tw, and the body of data simply is inadequate to evaluate the influence of
Tw/Taw on ıpƍ/qe.
Taken together, Figs. 2 and 3 suggest that the measured pressure fluctuation intensity is influenced more
strongly by the inherent frequency response of the sensor and associated electronics than any physical parameter
save Mach number. Thus, the observed data scatter in Fig. 1 is, to a great extent, the result of measurement
limitations rather than physical effects, and furthermore, the correct magnitude of ıpƍ/qe probably is biased towards
the high side of the data range where the sensor frequency response is largest. This implies that the Lowson [8] and
Laganelli [50] models will return too low a value, which is consistent with the earlier observation that their
incompressible value of ıpƍ/qe=0.006 appears too low given the evolution of incompressible experiments. In fairness
to prior experiments, many of the authors whose measurements are represented in Fig. 1 indeed recognized that
limits in spatial resolution and frequency response would bias their results low, and Fig. 3b illustrates this point.
Nevertheless, such data often have been employed anyway in developing predictive models, which serves to
motivate the rigorous experiments that form the core of the present document.
The same data also may be presented as normalized by the wall shear stress, IJw, for those experiments in which
IJw was measured or predicted. Figure 4 displays these results, from which it is evident that the data scatter is at least
as broad as in Fig. 2, and any trend with Mach number is more difficult to determine. This probably is because even
for a carefully conducted experiment, the measurement or prediction of IJw (or cf) is a greater challenge at high
speeds and/or thin boundary layers than it is for a well-crafted incompressible experiment. Farabee and Casarella
[21] studied the incompressible data for ıpƍ/IJw and determined a constant value of 2.55 for low Reynolds numbers,
rising slowly as Re increased. For the range of Reynolds numbers of the historical data shown in Fig. 4, Farabee and
Casarella’’s prediction would raise ıpƍ/IJw only to about 3.9; this range of values is shown in Fig. 4 as the
incompressible data bar. With the exception of Kistler and Chen [32], all the high-speed data is at or lower than this
range.
To assess whether the scatter in the ıpƍ/IJw data points follows any particular dependence, trends versus Reį, Reș,
gį/U’, and Uc/ʌd were examined analogously to Figs. 2 and 3. A meaningful trend could be found only with gį/U’,
which is shown in Fig. 5, and the other three plots have been omitted for brevity. Figure 5 supports the earlier
indication that the dominant influence upon the measured pressure fluctuation intensities is the frequency response
of the sensor, and that larger values of ıpƍ/IJw appear more likely to be accurate.

Experimental Apparatus
Trisonic Wind Tunnel
Experiments were performed in Sandia’’s Trisonic Wind Tunnel (TWT), which is a blowdown-to-atmosphere
facility using air as the test gas through a 305 × 305 mm2 (12 × 12 inch2) rectangular test section enclosed within a
pressurized plenum. In its supersonic configuration, the TWT’’s Mach numbers are varied by exchanging the top
and bottom walls of the planar nozzle; Machs 1.5, 2.0, 2.5, and 3.0 are available. The freestream Reynolds number
ranges from 24 × 106 m-1 to 44 × 106 m-1 in the current study, which covers most of the supersonic capability of the
TWT. The wind tunnel stagnation temperature T0 is fixed at 321K ± 2K by heating in the storage tanks, and the wall
temperature is effectively constant at ambient conditions, Tw=307K ± 3K, though Tw does tend to drift upward
within this range as the wall warms during the course of a day.
Wall pressure measurements were conducted at three different streamwise stations on the side wall of the TWT
by mounting them in a window blank as shown in the top portion of Fig. 6. A distance of 441 mm separates each
station. Though these locations are still within the nozzle curvature, the supersonic expansion is complete in the
vicinity of the centerline where the sensors are placed; i.e., the measurement stations lie within the ““test rhombus”” of
the test section as evidenced by wind tunnel calibrations. Pitot probe surveys of the boundary layer were conducted

-6-
American Institute of Aeronautics and Astronautics
at each of the three streamwise stations, then fitted to
the method of Sun and Childs [51] to provide the
freestream velocity U’, the 99%-velocity boundary
layer thickness į, the displacement thickness į*, and the
momentum thickness ș. The skin friction coefficient cf
also may be determined by the Sun and Childs fit.
Within the range of these measurement stations, the
boundary layer has been shown to be in equilibrium, as
discussed by White [52] based upon Clauser’’s classic
argument [53].

Pressure Instrumentation
Three different types of pressure sensors were
used: Kulite MIC-062, PCB 132A31, and Brüel &
Kjaer (B&K) 4138. Additionally, two Kulite screen
types were tested, and two of each sensor were
employed, bringing the total number of sensors to eight.
The present document utilizes only those sensors that
proved most successful; the others are discussed in [4].
The eight sensors were installed in the wind tunnel
simultaneously, mounted in an insert block that placed
all sensors at the same axial station within the test
section, as indicated in Fig. 6. The spanwise variation Fig. 6: Sketch of the three sensor stations within the
in sensor position is not significant, as determined by Trisonic Wind Tunnel, and the sensor mounting block
wind tunnel calibration data. The insert block that may be installed at any station.
containing the eight installed sensors could be placed at
any of the three streamwise stations, leaving the other
two locations to be plugged by a blank containing no sensors.
The Kulite MIC-062 is a piezoelectric silicon sensor based upon the widely used XCQ-062 product line and is
essentially a 7 kPa pressure transducer with a specified sensitivity of 2 Pa and a diaphragm resonance frequency of
approximately 125 kHz in a cylindrical housing of 1.62 mm diameter. Data reported herein used an A-screen to
protect the diaphragm, which is simply a 1.02 mm hole at the sensor face, as opposed to the more conventional B-
screen consisting of a ring of 0.15 mm holes around the periphery of the sensor face. The B-screen offers better
protection against damage and particulates, but its dynamic response is flat only to about 20% of the diaphragm
resonance frequency whereas the A-screen provides superior frequency response to about 30-40% [4, 54].
However, the geometry of the A-screen introduces cavity resonance at a much lower frequency. The Kulite
microphones are differential pressure transducers and have a vent tube extending from their rear, which was
plumbed to a surge tank whose pressure could be set prior to each wind tunnel run to a value within the sensor
range; venting to the tunnel plenum led to saturation of the pressure signal. Since the pressure signals are recorded
with no DC response (see below) to focus solely upon the fluctuations, it is not necessary to accurately measure the
surge tank pressure or even to carefully set this value.
The PCB 132A31 is a piezoelectric quartz sensor intended by the manufacturer to be a time-of-arrival sensor,
but has been used successfully to detect very high frequency events in hypersonic wind tunnel experiments [54-56].
Its performance at low frequencies is limited by a cutoff of 11 kHz but its high-frequency performance exceeds
1 MHz with a specified sensitivity of 7 Pa. The manufacturer suggests the high-frequency response is flat to
approximately one-third of its maximum, or roughly 300 kHz. The quartz crystal is bonded to the underside of the
sensing surface and hence does not require exposure to the flow through a screen as do Kulite sensors, preventing
the occurrence of a cavity resonance. The effective sensing area is stated by the manufacturer as 1.6 mm2.
Finally, the B&K 4138 is a pressure-field condenser microphone of diameter 3.2 mm ordinarily used as a
calibration standard with a sensitivity as low as 0.01 Pa. Of the sensors utilized in the present work, only the B&K
4138 is provided by the manufacturer with a calibration curve of amplitude response as a function of frequency up to
140 kHz. Its purpose in the current study is to function as a reference standard in sensor calibrations performed
acoustically outside of the wind tunnel.

Signal Processing
The Kulite sensors were supplied a 10 V excitation from an Endevco Model 136 Amplifier, which also provided

-7-
American Institute of Aeronautics and Astronautics
the first stage of signal gain for the sensor output. The Kulite signals then were low-pass filtered at 200 kHz by a
Krohn-Hite Model 3384 active filter, which is able to provide 48 dB attenuation per octave. Meanwhile, the
PCB132 sensors use a combined power supply and signal conditioner (PCB 482A22), then their signals were low-
pass filtered at 500 kHz by a Krohn-Hite Model 3944 active filter; though it only offers 24 dB attenuation per octave
in comparison to the Model 3384, it is capable of reaching significantly higher frequency cutoffs and hence is
necessary for the high-frequency signals of the PCB 132 sensors.
All signals, once amplified and filtered, were sampled using a National Instruments PXI-6133 analog-to-digital
converter housed in a PXI-1042 chassis. Sampling frequencies were 1 MHz, well below the 10 MHz bandwidth
provided by the PXI-6133 cards, and were recorded at 14 bits depth for typically 1 second of length. These A-to-D
converters have been shown to have excellent low-noise characteristics despite their large bandwidth and performed
better than other available sampling electronics.

Data Analysis
To properly analyze fluctuating pressure data, the sensors first must be dynamically calibrated to provide the
amplitude response as a function of frequency. The Kulite sensors are intended for use up to 40 kHz, a range over
which their frequency response should be approximately flat, and a conventional static calibration proved sufficient.
Conversely, the PCB sensors are intended for use above 20 kHz and should have a flat response to roughly 300 kHz,
but their amplitude response is poorly characterized and a static calibration is impossible due to their low-frequency
cutoff. A dynamic calibration similar to the one conducted by Zuckerwar et al [57] is used in the present case, with
the principal difference that Zuckerwar et al performed a free-field calibration for acoustic measurements whereas
the current implementation is a pressure-field calibration for wall-mounted sensors. The calibration is based upon
the substitution method, where the B&K 4138 reference standard is flush-mounted in the center of a nearly infinite
wall within an anechoic chamber, then subsequently replaced by each sensor to be calibrated. The sensors are
excited by an ultrasonic speaker reaching 45 kHz. By maintaining constant output of the speaker and constant
ambient conditions within the calibration site, the signal from the calibrated sensor may be compared to that from
the B&K reference and hence a calibration curve is produced.
Additionally, the low-frequency data from the Kulite sensor are corrected for the effects of wind tunnel noise
and vibration using an adaptive filtering scheme much like that described by Naguib et al. [5]. Two identical Kulite
sensors are installed into the wind tunnel separated by 8-12 boundary layer thicknesses. At this distance, the sensors
can be expected to return statistically identical signals but are sufficiently separated as to observe independent
turbulent events. Thus, the correlated portion of the two signals is due only to wind tunnel noise and the adaptive
filter yields a signal approximating the noise. By subtracting this noise signal from an original Kulite signal, a
noise-cancelled signal is produced and its power spectrum reflects the true turbulence of the flow. See [4] for
details. This procedure was implemented only on the Kulite A-screen sensors, since it is these that produce the
signal used for the low-frequency portion of the spectrum.
One of the great difficulties with measuring fluctuating wall pressures in high-speed flows is the small spatial
scale associated with the high-frequency fluctuations relative to the size of the sensors themselves. Corcos [6]
provides the most widely known analysis of the attenuation of pressure signals due to limited spatial resolution, in
which he characterizes the severity of the problem based upon the nondimensional sensor size, Ȧd/2Uc, where d is
the diameter of the sensor’’s sensitive area and Uc is the convective velocity. Corcos finds that a nondimensional
sensor size of unity represents the 3-dB point of spatial attenuation; hence, when Ȧd/2Uc < 1 minimal signal
attenuation occurs, and when Ȧd/2Uc > 1, signal attenuation becomes a concern. The Corcos correction has been
used widely for incompressible flows and found to return useful results [1, 16, 17], though with limits. Lueptow’’s
[18] analysis of a numerical simulation suggested that the Corcos correction is effective if roughly Ȧd/2Uc < 4,
which is consistent with Schewe’’s [15] data.
In the present case, the 3-dB damping point for the Kulite A-screen sensor is 140 kHz based on the 1.02 mm
cavity in the sensor face; given that this sensor is effective only to about 40 kHz, spatial attenuation does not present
a difficulty. Conversely, the 3-dB damping point for the PCB132 based on the 1.6 mm2 sensitive area is 100 kHz,
which poses a considerable problem given that this sensor has a frequency response reaching 1 MHz. (The sensitive
area of the PCB132 actually is a rectangle 1.0 mm × 1.6 mm, but its orientation is not marked on the sensor housing
and is indeterminate; satisfactory results have been obtained assuming a circular sensitive area of 1.6 mm2.) Corcos
[6] provides both a semi-empirical formula and tabulated values for the signal attenuation as a function of Ȧd/2Uc,
which then can be used to correct measured values. However, to further complicate the problem, Uc is a function of
Ȧ because high-frequencies typically result from small-scale turbulent eddies near the wall, which convect at a
slower velocity than larger eddies further from the wall. Therefore, the convection velocity must first be known as a
function of the pressure fluctuation frequency, which can be difficult to directly measure for the small spatial scales

-8-
American Institute of Aeronautics and Astronautics
at which it is most necessary. Corcos himself used a curve yielding Uc/U’=0.8 for low frequencies and asymptoting
to 0.6 at high frequencies. In the present case, an equation fit to his curve has been adjusted to return Uc/U’=0.85 at
low frequencies because cross-correlations between streamwise arrays of pressure sensors show this to be the correct
value at large wavelengths. Practically speaking, however, the Corcos correction is significant only at high
frequencies, where a constant Uc/U’=0.6 is a reasonable approximation. However, the Lueptow [18] and Schewe
[15] analyses limit the validity of the Corcos correction to 400 kHz for the PCB132; this represents the effective
high-frequency limit of measurements in the present study.
Once the data from the Kulite A-screen and PCB132 sensors have been calibrated, filtered, and corrected for
spatial attenuation, they agree with each other quite well in their overlap range from about 20-40 kHz –– in fact, their
difference lies within the measurement uncertainty [4]. To create a spectrum ranging from the lowest to highest
measurable frequencies, these two signals can be merged by weighted averaging in the 20-40 kHz range, then using
the Kulite-A data below 20 kHz and the PCB132 data above 40 kHz. The result is a spectrum spanning the
frequency range from 100 Hz to 400 kHz; the low frequency limit is determined by practical data processing
restrictions, and the high frequency limit is determined by the validity of the Corcos correction. This procedure
yields spectra over a much larger frequency range than either sensor can provide alone, and it is explained in detail
in [4].

Results and Discussion


Data have been acquired at all three streamwise stations for six different values of freestream Reynolds numbers
at all four supersonic Mach numbers within the capability of the TWT. The complementary Pitot probe surveys at
identical conditions and stations provide the boundary layer thickness, the displacement thickness, and the
momentum thickness. Inner boundary layer variables have been determined based upon the skin friction coefficient
found from the Sun and Childs [51] fit. Data have been acquired within a range of wind tunnel unit Reynolds
numbers of Re’=24 - 44 × 106 m-1 for Machs 1.5, 2.0, and 2.5, and Re’=32 - 44 × 106 m-1 for Mach 3.0 because the
TWT will not start at lower values. This corresponds to values of Reș ranging from 24000 to 51000, depending
upon station and Mach number. Run conditions are summarized in Table 1.

Pressure Power Spectra


Representative examples of the power spectral density of the pressure fluctuations are shown in Fig. 7 for each
of the four Mach numbers. Data are provided at the downstream station for a mid-range Reș of about 35000,
varying by ±1000 between the Mach numbers. Data are given as a function of the angular frequency Ȧ and have
been normalized based upon the freestream velocity U’, the freestream dynamic pressure q’, and the boundary layer
displacement thickness į*. The pressure spectra clearly show that an increase in Mach number produces a reduction

Table 1: Experimental parameters.

Mach P0 (kPa) U’ (m/s) station į (mm) į* (mm) ș (mm) cf (×10-3) Reș


1.47 182 - 334 441 upstream 12.0 - 12.5 1.88 - 2.04 0.89 - 0.95 1.80 - 1.98 23000 - 39000
middle 13.9 - 14.4 2.16 - 2.34 1.02 - 1.09 1.76 - 1.92 26000 - 45000
downstream 16.0 - 16.5 2.45 - 2.63 1.16 - 1.24 1.73 - 1.88 30000 - 51000
1.98 217 - 398 537 upstream 10.0 - 10.3 1.81 - 1.97 0.64 - 0.69 1.77 - 1.93 17000 - 28000
middle 11.8 - 12.0 2.28 - 2.49 0.81 - 0.87 1.65 - 17.8 21000 - 36000
downstream 13.6 - 14.3 2.68 - 2.91 0.96 - 1.02 1.58 - 1.73 24000 - 42000
2.50 277 - 507 603 upstream 10.3 - 10.4 2.26 - 2.45 0.61 - 0.64 16.0 - 17.3 15000 - 27000
middle 12.4 - 12.5 2.81 - 3.06 0.76 - 0.80 15.1 - 16.4 19000 - 33000
downstream 14.7 - 15.0 3.46 - 3.76 0.94 - 0.99 14.5 - 15.5 24000 - 41000
3.02 478 - 660 649 upstream 11.3 - 11.4 3.13 - 3.26 0.64 - 0.66 13.8 - 14.4 21000 - 28000
middle 13.7 - 13.9 3.92 - 4.12 0.80 - 0.83 13.0 - 13.4 27000 - 35000
downstream 16.8 - 16.9 4.78 - 4.93 0.98 - 1.04 12.5 - 12.8 33000 - 43000

-9-
American Institute of Aeronautics and Astronautics
Fig. 7: Power spectral densities of the wall pressure fluctuations at Reș=35000 and the
downstream station. The light blue line shows data for Mach 2.0 prior to correction for spatial
attenuation. The thin black lines show data prior to compensation for wind tunnel noise.

in the normalized magnitude, though the shape of the spectra remains similar, which is consistent with the data
compiled by Laganelli et al [50]. The normalized frequency at which the rolloff becomes appreciable,
approximately a value of 1, also is consistent with the spectra reported in [50]. In the Mach 2.5 spectrum, a small
spike may be seen near Ȧį*/U’ = 0.6, which probably is attributable to an unsteady Mach wave disturbance residual
from the nozzle expansion.
Additionally plotted in Fig. 7 is the slope defined by Ȧ-1. Bradshaw [58] predicted that the lower frequency
range of the pressure spectra is induced by eddies in the logarithmic layer of the boundary layer and should exhibit a
Ȧ-1 dependence. Panton and Linebarger’’s [59] and Blake’’s [60] analyses each draw the same conclusion, but some
measurements have suggested that the slope actually is a little less than -1 if the Reynolds number is not sufficiently
large [19, 61]. In the present case, the Corcos-corrected spectra indicate a slope very close to Ȧ-1 in the region where
the logarithmic layer is influential. The consistency of the present data with the predicted slope bolsters the
reliability of the present measurements and the approach taken to merge sensor signals. The light blue curve in
Fig. 7 is the power spectrum for Mach 2.0 prior to the implementation of the Corcos correction for spatial
attenuation, which highlights the large magnitude of the correction at high frequencies. Unfortunately, the
electronic noise present at high frequencies becomes amplified as well, but this is an unavoidable consequence of an
otherwise successful application of the Corcos correction. It is evident that the Ȧ-1 slope cannot be detected without
the use of the Corcos correction.
The thin black lines in Fig. 7 are the original noisy power spectra prior to the use of the adaptive filter for wind
tunnel noise cancellation. Clearly, the noise effects are relatively mild in comparison to the magnitude of the
boundary layer fluctuations and are constrained to frequencies below 1 kHz (and mostly below 500 Hz). The Mach
1.5 and 2.0 data show a small but noteworthy reduction in amplitude once noise cancellation is implemented, but no
appreciable effect is found for Mach 2.5 and 3.0.
Another significant conclusion from the power spectra in Fig. 7 may be observed near the low-frequency limit.
The low-frequency behavior of boundary layer wall pressure spectra has proven to be somewhat controversial, with
some theories predicting a Ȧ2 dependence whereas others propose a slower decrease in the power as frequency
drops, or even a flat response (see Bull’’s summary in [17]). Experimentally, it is especially difficult to obtain
reliable measurements in this frequency range owing to the influence of facility noise and the limits of
instrumentation. Of experiments that employed a noise-cancellation scheme, it would appear that only Farabee and
Casarella [21] have demonstrated the existence of the Ȧ2 regime; however, other such experiments [19, 20] that do

-10-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 8: Power spectra for Mach 2.0 at the downstream station for all six values of Reș; (a) absolute units; (b)
normalized units.

not support the Ȧ2 dependence do not reach as low a normalized frequency as do Farabee and Casarella.
Significantly, glider experiments have returned mixed results [16, 62, 63] and atmospheric boundary layer
measurements [64] do not find the Ȧ2 regime, but again, these may not explore a sufficiently low frequency range.
In the present case, the data clearly show no Ȧ2 dependence at low frequencies, instead indicating a flat or slightly
decreasing spectra except at Mach 1.5. A comparison of noise-cancelled signals to those prior to the adaptive
filtering algorithm shows that only a small portion of the power spectrum amplitude is attributable to wind tunnel
noise, suggesting that a more finely tuned noise-cancellation filter would not be beneficial; similar results were
found when instead employing a subtractive noise-cancellation scheme like that of Simpson et al [65].
The measurement uncertainty is composed of the data repeatability and biases in the calibration and data
corrections. Three repeat runs were conducted for the Mach 2.0 conditions, providing a total of eight redundant data
sets since two sets of all the sensors were employed in each case. The uncertainty due to repeatability, found as the
95% confidence interval, was determined from the scatter between these eight cases as 6 × 10-6, in normalized units.
The scatter was computed using a narrow rolling average across the spectra to remove the effects of the noise
between adjacent frequency bins, which shows that the difference between repeated measurements is quite small in
comparison to the routine noise in computing the spectra. The bias error principally lies in the sensor calibrations.
Based upon repeated calibrations using the ultrasonic speaker, a range of calibration coefficients was established and
the extreme values used to re-reduce the data. This yielded a bias error of about 6 × 10-6, coincidentally the same as
the repeatability. Combining the two error sources leads to an uncertainty estimate of 8 × 10-6, which is the value
plotted as an error bar in Fig. 7. Of course, additional uncertainty exists at higher frequencies concerning the
veracity of the Corcos correction. Lueptow’’s work [18] suggests that the error will be considerably less than the 8 ×
10-6 value already cited until the limit of the Corcos correction is approached, and hence can be neglected below the
highest measured frequencies, at which point the electronic noise becomes a greater concern anyway.
Data acquired at Mach 2.0 in the downstream station are shown in Fig. 8 for all six Reynolds numbers, where
Fig. 8a provides the results in absolute units and Fig. 8b shows them normalized. Even for this relatively mild
variation in Reș, the gradual increase in magnitude of the power spectra is visible in Fig. 8a. However, once
normalized, all six cases collapse neatly into one curve. Results at other Mach numbers and streamwise stations are
similar. Spectra at the three streamwise stations are given in Fig. 9, again with absolute units in Fig. 9a and
normalized units in Fig. 9b. The boundary layer thickness į changes from 10.0 mm to 12.0 mm to 14.3 mm, from
upstream to middle to downstream stations, respectively. As the boundary layer thickens, the power spectrum
diminishes in amplitude slightly and shows the beginning of rolloff to the Ȧ-1 slope at a lower frequency. Such
behavior may be attributed to the decrease in dynamic pressure required to maintain a constant Reynolds number
while the boundary layer thickness rises. The strength of the pressure fluctuations is proportional to the dynamic
pressure, so this change in freestream conditions is reflected in the power spectra. However, as Fig. 9b makes clear,
the three power spectra appear to be identical in normalized form. Although differences near the high-frequency
limit are difficult to ascertain because of the prominence of sensor noise, the three curves collapse well throughout
the measurable spectrum. This successful normalization provides supporting evidence that the wall boundary layer

-11-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 9: Power spectra for Mach 2.0 at the three streamwise stations for Reș=28000; (a) absolute units; (b)
normalized units.

(a) (b)
Fig. 10: Power spectra for Mach 2.0 normalized on inner variables; (a) varying Reș at the downstream station; (b)
all three streamwise stations for constant Reș=28000.

under investigation is in equilibrium.


uilibrium. Once appropriately normalized, the Reynolds number, freestream conditions,
and streamwise station cease to influence the power spectra within the measurable frequency range; the only
parameter with an effect is the freestream Mach number.
The normalizations used in Figs. 8 and 9 are based upon outer flow variables, but have been quite successful to
the highest frequencies measurable in the present experiment, though these are considered to be mid-range
frequencies within the hierarchy of boundary layer turbulence. Another nondimensionalization is based upon inner
flow variables and is shown in Fig. 10, where Fig. 10a shows inner variable normalization of the Mach 2.0 data at
multiple Reynolds numbers from Fig. 8 and Fig. 10b shows the analogous results using the Mach 2.0 data from the
three streamwise stations from Fig. 9. Figure 10a shows some degree of greater spread at low frequencies than the
outer variable scaling of Figs. 8b, which is as expected, but the collapse of the curves at higher measured
frequencies appears equally good using either approach. In contrast, Fig. 10b shows as good collapse across the
entire measured spectrum for inner variables as for outer variables in Fig. 9b. The differing behavior of Fig. 10b
compared with Fig. 10a may be because the Reynolds number is held constant in the data of Fig. 10b whereas it
varies for the data of Fig. 10a.

Pressure Fluctuation Intensity


The intensity of the measured pressure fluctuations can be reduced to a single defining parameter as the r.m.s. of
the unsteady signal, ıpƍ, just as in the historical data of Fig. 1. Ordinarily, this value simply is calculated from the

-12-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 11: Pressure fluctuation magnitude as a function of Mach number. Filled gray points show the current
experiment, colored points and lines show the historical database as in Figs. 1 and 4. Data are normalized by: (a)
dynamic pressure; (b) wall shear stress.

time-series pressure signal, but in the present case of merged data from multiple sensors combined with the Corcos
correction, ıpƍ is found by integrating the power spectral density curves. Figure 11 plots the present results against
the historical data of Figs. 1 and 4, using data at Reș=35000 and the downstream station as representative of the
larger data set. The data are shown normalized by qe in Fig. 11a and by IJw in Fig 11b. The current measurements
are shown in three capacities; in one case, the corrected spectra (gradient triangles) are used to compute ıpƍ, and in a
second case (delta triangles), the originally recorded data are used instead, without the Corcos corrections or the
noise cancellation. The third set of points (diamonds) will be explained subsequently. The corrected data yield
results somewhat in excess of the bulk of the historical data, but the uncorrected data fall towards the lower end of
its wide scatter; this is true whether normalized by freestream dynamic pressure or wall shear stress. Few of the
historical experiments have employed the Corcos correction that has proven key to providing correct power spectra
at high frequencies in the present work. Given that the current measurements agree fairly well with the bulk of the
historical data when no corrections are used but rise appreciably above it with corrections, it would appear that much
of the historical database for supersonic and hypersonic pressure fluctuations may be biased low. The same
conclusion is evident regardless of normalization parameter.
The nondimensional frequency of the measurements in TWT, gį/U’, was calculated using the cutoff frequency
due to spatial resolution, g=Uc/ʌd. This must be done twice, once for the uncorrected TWT measurements and again
for the Corcos-corrected measurements. Figure 12 displays the results superposed over the corresponding values
from Figs. 3a and 5 for the historical database. The current frequency limitations well fit those of the historical data,
with the corrected data points possessing nondimensional frequencies equivalent to other experiments returning the
largest values of pressure fluctuation intensity. Similarly, the reduced values of the nondimensional frequencies for
the uncorrected TWT data points are close to those of the historical data clustered nearby. These observations are
true whether ıpƍ is normalized by qe or IJw. The consistency of the TWT results, both the values of ıpƍ and those of
the frequency response gį/U’ used to obtain them, with those of the historical data provide further support for the
notion that the largest values of ıpƍ are more likely to be correct and the bulk of the historical data is biased low ––
just as the limited frequency response of the uncorrected TWT data bias low their values of ıpƍ.
Even with the data corrections and measurements extending to 400 kHz, the full range of energy-containing
turbulent eddies has not been captured by the present measurements. Somewhere beyond the Ȧ-1 dependence of the
logarithmic layer, the spectra can be expected to transition to a steeper slope and become proportional to Ȧ-5 as small
wavelength eddies very near the wall and viscosity effects become dominant [60]. This high-frequency region has
been confirmed in such incompressible experiments as [19, 20] and its onset can be expected at about ȦȞ/uIJ2=0.3
[21], which translates to about g=350-500 kHz in the current experiment depending upon conditions. The spectrum

-13-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 12: Data points shaded by sensor non-dimensional cutoff frequency, normalized by: (a) dynamic pressure; (b)
wall shear stress. Circles represent the historical database as from Figs. 3a and 5, delta triangles the TWT
uncorrected data points from Fig. 11a, and gradient triangles the TWT corrected data points from Fig. 11b.

is believed to pass through a region proportional to Ȧ-7/3


as it transitions from Ȧ-1 to Ȧ-5 [15, 66], fully reaching
the Ȧ-5 dependency at about ȦȞ/uIJ2=1.0 [60]. Effects of
the high-frequency portion of the spectrum have been
appended to the present data by extending the
logarithmic region to its limit at ȦȞ/uIJ2=0.3, then
following the Ȧ-7/3 dependence until ȦȞ/uIJ2=1, at which
point the Ȧ-5 dependence is obeyed until the spectra has
fallen to such low levels that it contains no significant
energy. This extended spectrum is sketched in Fig. 13
for the Mach 2 case. Integration over the augmented
curve yields a value for ıpƍ that estimates the increased
intensity when the entire spectrum is accounted; these
values are shown in Fig. 11 as the ““TWT extended””
data points (diamonds). The integrated value of the Fig. 13: Data for Mach 2 extended to higher
estimated high-frequency region is within about 10% of frequencies by estimating the power spectrum shape.
that calculated by Farabee and Casarella [21], which
lends some confidence to the utility of this exercise.
Clearly, Fig. 11 shows that additional energy is contained in the immeasurable portion of the spectrum,
providing further evidence that the historical database is biased low from the true value of pressure fluctuation
intensity. The exact values of the extended spectra are no better than an estimate, since they extrapolate from
generalized behavior found for incompressible flows rather than providing any new data at such extraordinarily
large frequencies, but they do indicate that substantial energy is missed by any achievable supersonic measurement.
However, in many practical cases for flight vehicle design, the neglected fluctuation energy may be of minimal
concern. Between the Corcos correction and the unmeasured Ȧ-5 region, the great majority of the lost energy lies at
very high frequencies. Most fluid-structure interactions occur at considerably lower frequencies and the linear, or
nearly so, structural response of most vehicle components indicates that the behavior at the lower portions of the
spectrum is of greater importance. Conversely, other flight applications, particularly those motivated by
aeroacoustics, are concerned with much higher frequency events, and therefore accurate prediction across the entire
spectrum may be more relevant.
The dependence of the measured pressure fluctuation intensities on Reynolds number was examined to compare
with the limited results from the compressible historical database. Figure 14 shows trends against Reș for both ıpƍ/qe
and ıpƍ/IJw, given in Figs. 14a and 14b respectively using the corrected power spectra. Values of ıpƍ resulting from

-14-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 14: Pressure fluctuation intensity as a function of momentum thickness Reynolds number, normalized by: (a)
dynamic pressure; (b) wall shear stress. Delta triangles represent the upstream station, squares the middle station,
and gradient triangles the downstream station. Power-law curve fits also are shown.

the extended spectra are not considered because these are approximate estimates based upon theory and
incompressible experiments rather than measured data. Data from the three streamwise stations are shown using
different symbols, from which it can be seen that the three stations collapse well into one curve for each Mach
number when ıpƍ is normalized by qe. At Mach 1.5, the pressure fluctuation intensity showsshows a very slight increase
with Reynolds number, but at the other three Mach numbers, clear decreasing trends are seen consistent with
previous high-speed studies [32, 33, 43, 44]. A power-law curve fit to each Mach number also is provided, from
which a dependency for the three larger Mach numbers of about ıpƍ/qe ~ Reș-0.1 emerges.
When scaled by IJw in Fig. 14b, the results become considerably more jumbled. More scatter is observed
between points acquired at different streamwise stations and three of the four Mach numbers exhibit significant
overlap. The contrast of this variability with the smooth and consistent curves of Fig. 14a may indicate difficulties
in the measurement of IJw from the Pitot probe surveys in comparison to the easily obtained values of q’.
Nevertheless, the data show a clear trend of rising ıpƍ/IJw as Reș increases for all Mach numbers, though the strength
of the trend varies. Differences in the IJw-normalized results at each streamwise station may contribute to the
strength of the rising trend in ıpƍ/IJw.

Comparison to Predictive Models


During the many decades in which wall pressure spectra have been measured and analyzed, a number of semi-
empirical engineering models have been developed. Such models may be compared to the present data as a means
of assessing their robustness and applicability to the current flow conditions. To date, the present data have been
compared to the Maestrello [33] model that was derived from supersonic wind tunnel wall measurements, the
Laganelli [50] model designed for supersonic and hypersonic attached boundary layers, the Efimtsov model [67]
based on aircraft wind tunnel testing, and the Goody [68] model for a broad range of Reynolds numbers in
incompressible boundary layers. The Goody model is of interest despite its lack of compressibility effects because
Hwang et al’’s [69] recent review of several models finds it demonstrates the best performance, perhaps in part due
to its recent pedigree.
Figure 15 shows the present data at all four Mach numbers compared to the aforementioned models. The data
are shown for the downstream station and a common Reș=35000 (same as Fig. 7), but given the normalized
parameters, any of the data acquired in the present work will collapse to one of these four curves depending upon
Mach number. A common trend immediately identifiable in Fig. 15 is that every model returns amplitudes
considerably less than the measurements. For the Maestrello, Laganelli, and Efimtsov models, which are based
upon experimental data in the compressible regime, this may be a reflection of the low magnitudes returned for
much of the historical database in comparison to the present work. The Goody model, on the other hand, is built
upon more recent data believed to exhibit greater accuracy, so instead may be due to differences between the
incompressible and compressible flow regimes. The Laganelli and Efimtsov models do not return a Ȧ2 dependence,

-15-
American Institute of Aeronautics and Astronautics
(a) (b)
Fig. 15: Comparison of current data from Fig. 7 to engineering models; (a) Maestrello [33] and Laganelli [50]
models; (b) Efimtsov [67] and Goody models [68].

by design, which matches the present experimental


observations; if a Ȧ2 dependence does exist, the
Maestrello and Goody models predict it to reach too
high a frequency. The Maestrello and Laganelli models
do not display an appropriate Ȧ-1 trend at the highest
measurable frequencies, but the Efimtsov and Goody
models appear to come close. The Goody model
actually is designed to reach a Ȧ-0.7 dependence for
sufficiently large Reynolds number, which is consistent
with some incompressible data [68]. Finally, all models
show the correct trend of a reduced fluctuation
magnitude as the Mach number increases, although they
may not well predict the spread in magnitudes as the
Mach number varies (most notably, Laganelli).
The Laganelli model was developed assuming the Fig. 16: Comparison of current data from Fig. 7 to the
incompressible pressure fluctuation intensity of modified Laganelli and Goody models.
ıpƍ/qe=0.006, which now is believed to be more
accurately stated as ıpƍ/qe=0.009. Laganelli himself
recognized the evolution of the accepted incompressible value and used ıpƍ/qe=0.0078 in a subsequent model [3].
Laganelli’’s original model [50] is easily adjusted to incorporate the ıpƍ/qe=0.009 value and is shown in Fig. 16.
Similarly, a compressible adjustment to Goody’’s model [68] may be attempted by introducing Laganelli’’s
compressible correction factor to the magnitude of the incompressible skin friction coefficient. These two modified
predictions are displayed in Fig. 16, demonstrating that each returns results nearer the data than their original
formulations, but still do not accurately reflect the measurements. The magnitudes of the modified Laganelli
predictions reside nearer the group average at low frequencies, but still fail to exhibit the spread between Mach
numbers found in the data. The modified Goody model continues to return magnitudes well below that of the
present data.
Clearly, none of the predictive models under consideration returns robust and accurate results in comparison to
the present measurements. This may be partly explained by their basis in historical measurements likely to have
been biased towards lower magnitudes and suffered from frequency response limitations. Further development of
such models for compressible flows is warranted.

Conclusions
Accurate measurement of fluctuating wall pressure spectra beneath a supersonic turbulent boundary layer has
proven elusive, such that a compilation of past efforts exhibits an alarming degree of scatter and hinders the
development of engineering models. Recent experiments conducted in Sandia’’s Trisonic Wind Tunnel up to Mach 3

-16-
American Institute of Aeronautics and Astronautics
have provided wall pressure data to frequencies reaching 400 kHz to help reconcile conflicts in the historical data
and promote the reliability of predictive modeling and simulation. Data were acquired using piezoresistive silicon
pressure transducers effective at low- and mid-range frequencies, then supplemented by piezoelectric quartz sensors
capable of detecting very high frequency events. The two sensor types were calibrated based upon a condenser
microphone reference standard, then their resulting data combined into a single curve describing the wall pressure
spectra. Data were corrected for spatial attenuation at high frequencies using the Corcos correction and for wind-
tunnel noise and vibration at low frequencies using a noise cancellation algorithm based upon adaptive filtering.
The resulting data revealed the Ȧ-1 dependence for fluctuations within the logarithmic region of the boundary
layer, as predicted by theory and found in some incompressible experiments. The power spectra are essentially flat
at low frequency and do not exhibit the theorized Ȧ2 dependence; measurements reached sufficiently low
frequencies that this region should have been evident were it existent. Variations in the Reynolds number of the
flow or the streamwise location of the measurement are observable in absolute units but collapse to a single curve
for each Mach number when normalized by outer flow variables. Normalization by inner flow variables is
successful for the Ȧ-1 region but less so for lower frequencies.
The overall magnitude of the pressure fluctuation intensity was calculated by integrating each power spectrum
and comparing the results to the historical database of compressible boundary layer measurements. A power-law
dependence was found of decreasing intensity as the Reynolds number was raised when normalized by the
freestream dynamic pressure, but increasing intensity when normalized by the wall shear stress. Data employing the
Corcos correction yield results somewhat in excess of the bulk of the historical data, but the uncorrected data fall
towards the lower end of its wide scatter; this is true whether normalized by freestream dynamic pressure or wall
shear stress. The effect becomes considerably greater when an estimate is made for the very high frequency
fluctuations beyond the reach of the measurement capability. A detailed examination of fifty years of historical data
reveals that the reported magnitude of the pressure fluctuation intensity chiefly is a function of the frequency
response of the sensors used. When the present results are considered without the Corcos correction, the resulting
frequency response is consistent with prior experiments that produced similarly low pressure fluctuation intensities;
conversely, when the Corcos correction is used, the frequency response matches those experiments that returned
elevated pressure fluctuation intensities. These trends suggest that much of the historical database for supersonic
and hypersonic pressure fluctuations may be biased low. Several semi-empirical predictive models have been
developed based upon historical data subject to these measurement limitations, but none of those considered in the
present study accurately represents the power spectra acquired during the present experiments.

Acknowledgments
The authors would like to thank Ryan Bond, Larry DeChant, Rich Field, Keith Miller, Jeff Payne, and Jerry
Rouse for numerous invaluable conversations regarding the physics of pressure fluctuations relevant to re-entry
vehicles. The compilation of the historical database was begun by Fred Blottner, now retired from Sandia, and the
authors are grateful for his contribution. He also developed implementations of three of the four power spectra
predictive models used herein. Tom Grasser designed much of the mounting hardware for the pressure sensors.

References
[1] Dolling, D. S., and Dussauge, J. P., ““Fluctuating Wall Pressure Measurements,”” AGARD-AG-315: A Survey of
Measurements and Measuring Techniques in Rapidly Distorted Compressible Turbulent Boundary Layers, May 1989, pp. 8-
1 to 8-18.
[2] Dolling, D. S., and Narlo, J. C., ““Driving Mechanism of Unsteady Separation Shock Motion in Hypersonic Interactive
Flow,”” AGARD-CP-428: Aerodynamics of Hypersonic Lifting Vehicles, April 1987.
[3] Laganelli, A. L., and Hinrichsen, R. L., ““Surface Pressure Fluctuation Correlations for High-Speed Flow,”” AIAA Paper
2000-2039, June 2000.
[4] Beresh, S. J., Henfling, J. F., Spillers, R. W., and Pruett, B. O. M., ““Measurement of Fluctuating Wall Pressures Beneath a
Supersonic Turbulent Boundary Layer,”” AIAA Paper 2010-0305, January 2010.
[5] Naguib, A. M., Gravante, S. P., and Wark, C. E., ““Extraction of Turbulent Wall-Pressure Time-Series using an Optimal
Filtering Scheme,”” Experiments in Fluids, Vol. 22, No. 1, 1996, pp. 14-22.
[6] Corcos, G. M., ““Resolution of Pressure in Turbulence,”” Journal of the Acoustical Society of America, Vo.. 35, No. 2, 1963,
pp. 192-199.
[7] Houbolt, J. C., ““On the Estimation of Pressure Fluctuations in Boundary Layers and Wakes,”” General Electric Report TIS
66SD296, April 1966.
[8] Lowson, M. V., ““Prediction of Boundary Layer Pressure Fluctuations,”” Air Force Flight Dynamics Laboratory Technical
Report AFFDL-TR-67-167, April 1968.

-17-
American Institute of Aeronautics and Astronautics
[9] Laganelli, A. L., Martellucci, A., and Shaw, L., ““Prediction of Surface Pressure Fluctuations in Hypersonic Turbulent
Boundary Layers,”” AIAA Paper 76-409, July 1976.
[10] Willmarth, W. W., and Wooldridge, C. E., ““Measurements of the Fluctuating Pressure at the Wall Beneath a Thick
Turbulent Boundary Layer,”” Journal of Fluid Mechanics, Vol. 14, 1962, pp. 187-210.
[11] Willmarth, W. W., and Roos, F. W., ““Resolution and Structure of the Wall Pressure Field Beneath a Turbulent Boundary
Layer,”” Journal of Fluid Mechanics, Vol. 22, Part 1, 1965, pp. 81-84.
[12] Bull, M. K., ““Wall-Pressure Fluctuations Associated with Subsonic Turbulent Boundary Layer Flow,”” Journal of Fluid
Mechanics, Vol. 28, Part 4, 1967, pp. 719-754.
[13] Shattuck, R. D., ““Sound Pressures and Correlations of Noise on the Fuselage of a Jet Aircraft in Flight,”” NASA TN-D-1086,
1961.
[14] Bull, M. K., and Thomas, A. S. W., ““High Frequency Wall-Pressure Fluctuations in Turbulent Boundary Layers,”” Physics of
Fluids, Vol. 19, No. 4, 1976, pp. 597-599.
[15] Schewe, G., ““On the Structure and Resolution of Wall-Pressure Fluctuations Associated with Turbulent Boundary-Layer
Flow,”” Journal of Fluid Mechanics, Vol. 134, 1983, pp. 311-328.
[16] Willmarth, W. W., ““Pressure Fluctuations Beneath Turbulent Boundary Layers,”” Annual Review of Fluid Mechanics, Vol. 7,
1975, pp. 13-38.
[17] Bull, M. K., ““Wall-Pressure Fluctuations Beneath Turbulent Boundary Layers: Some Reflections on Forty Years of
Research,”” Journal of Sound and Vibration, Vol. 190, No. 3, 1996, pp. 299-315.
[18] Lueptow, R. M., ““Transducer Resolution and the Turbulent Wall Pressure Spectrum,”” Journal of the Acoustical Society of
America, Vol. 97, No. 1, 1995, pp. 370-378.
[19] Gravante, S. P., Naguib, A. M., Wark, C. E., and Nagib, H. M., ““Characterization of the Pressure Fluctuations Under a Fully
Developed Turbulent Boundary Layer,”” AIAA Journal, Vol. 36, No. 10, 1998, pp. 1808-1816.
[20] Goody, M. C., and Simpson, R. L., ““Surface Pressure Fluctuations Beneath Two- and Three-Dimensional Turbulent
Boundary Layers,”” AIAA Journal, Vol. 38, No. 10, 2000, pp. 1822-1831.
[21] Farabee, T. M., and Casarella, M. J., ““Spectral Features of Wall Pressure Fluctuations Beneath Turbulent Boundary Layers,””
Physics of Fluids A, Vol. 3, No. 10, 1991, pp. 2410-2420.
[22] Chung, K. M., and Lu, F. K., ““An Experimental Study of a Cold-Wall Hypersonic Boundary Layer,”” AIAA Paper 92-0312,
January 1992.
[23] Chyu, W. J., and Hanly, R. D., ““Power and Cross-Spectra and Space-Time Correlations of Surface Fluctuating Pressures at
Mach Numbers Between 1.6 and 2.5,”” AIAA Paper 68-77, January 1968.
[24] Coe, C. F., Chyu, W. J., and Dods, J. B. Jr., ““Pressure Fluctuations Underlying Attached and Separated Supersonic
Turbulent Boundary Layers and Shock Waves,”” AIAA Paper 73-996, October 1973.
[25] Coe, C. F., ““Surface-Pressure Fluctuations Associated with Aerodynamic Noise,”” Basic Aerodynamic Noise Research,
NASA Report SP-207, 1969, pp. 409-424.
[26] Harvey, W. D., Bushnell, D. M., and Beckwith, I. E., ““Fluctuating Properties of Turbulent Boundary Layers for Mach
Numbers Up to 9,”” NASA TN D-5496, 1969.
[27] Heller, H. H., and Clemente, A. R., ““Unsteady Aerodynamic Loads on Slender Cones at Free-Stream Mach Numbers from 0
to 22,”” AIAA Paper 73-998, October 1973.
[28] Heller, H. H., and Clemente, A. R., ““Fluctuating Surface-Pressure Characteristics on Slender Cones in Subsonic,
Supersonic, and Hypersonic Mach Number Flow,”” NASA CR-2449, 1974.
[29] Howe, J. R., and Laganelli, A. L., ““Surface Pressure Fluctuation Measurements in Attached Transitional/Turbulent
Boundary Layers at Supersonic and Hypersonic Speeds,”” AIAA Paper 77-113, January 1977.
[30] Laganelli, A. L., and Howe, J. R., ““Prediction of Pressure Fluctuations Associated with Maneuvering Re-Entry Weapons,””
Air Force Report AFFDL-TR-77-59, Vol. 1, 1977.
[31] Garg, S., and Settles, G. S., ““Unsteady Pressure Loads Generated by Swept-Shock-Wave/Boundary-Layer Interactions,””
AIAA Journal, Vol. 34, No. 6, 1996, pp. 1174-1181.
[32] Kistler, A. L., and Chen, W. S., ““The Fluctuating Pressure Field in a Supersonic Turbulent Boundary Layer,”” Journal of
Fluid Mechanics, Vol. 16, No. 1, 1963, pp. 41-64.
[33] Maestrello, L., ““Radiation from and Panel Response to a Supersonic Turbulent Boundary Layer,”” Journal of Sound and
Vibration, Vol. 10, No. 2 1969, pp. 261-295.
[34] Martelucci, A., Chaump, L., and Rogers, D., ““Experimental Determination of the Aeroacoustic Environment about a Slender
Cone,”” AIAA Journal, Vol. 11, No. 5, pp. 635-642, 1973.
[35] Chaump, L. E., Martellucci, A., and Monfort, A., ““Aerodynamic Loads Associated with High Beta Re-Entry Vehicles,”” Air
Force Report AFFDL-TR-72-138, Vol. 1, 1973.
[36] Murphy, J. S., Bies, D. A., Speaker, W. V., and Franken, P. A., ““Wind Tunnel Investigation of Turbulent Boundary Layer
Noise as Related to Design Criteria for High Performance Vehicles,”” NASA TN-D-2247, 1964.
[37] Dolling, D. S., and Bogdonoff, S. M., ““An Experimental Investigation of the Unsteady Behavior of Blunt Fin-Induced
Shock Wave Turbulent Boundary Layer Interactions,”” AIAA Paper 81-1287, June 1981.
[38] Dolling, D. S., and Murphy, M. T., ““Unsteadiness of the Separation Shock Wave Structure in a Supersonic Compression
Ramp Flowfield,”” AIAA Journal, Vol. 21, No. 12, 1983, pp. 1628-1634.
[39] Dolling, D. S., and Or, C. T., ““Unsteadiness of the Shock Wave Structure in Attached and Separated Compression Ramp
Flows,”” Experiments in Fluids, Vol. 3, No. 1, 1985, pp. 24-32.

-18-
American Institute of Aeronautics and Astronautics
[40] Tan, D. K. M., Tran, T. T., and Bogdonoff, S. M., ““Wall Pressure Fluctuations in a Three-Dimensional Shock-Wave /
Turbulent Boundary Interaction,”” AIAA Journal, Vol. 25, No. 1, 1987, pp. 14-21.
[41] Muck, K. C., Andreopoulos, J., and Dussauge, J. P., ““Unsteady Nature of Shock-Wave/Turbulent Boundary-Layer
Interaction,”” AIAA Journal, Vol. 26, No. 2, 1988, pp. 179-187.
[42] Ringuette, M. J., Bookey, P., Wyckham, C., and Smits, A. J., ““Experimental Study of a Mach 3 Compression Ramp
Interaction at Reș=2400,”” AIAA Journal, Vol. 47, No. 2, 2009, pp. 373-385.
[43] Raman, K. R., ““Surface Pressure Fluctuations in Hypersonic Turbulent Boundary Layers,”” AIAA Paper 73-997, October
1973.
[44] Raman, K. R., ““A Study of Surface Pressure Fluctuations in Hypersonic Turbulent Boundary Layers,”” NASA CR-2386,
1974.
[45] Speaker, W. V., and Ailman, C. M., ““Spectra and Space-Time Correlations of the Fluctuating Pressures at a Wall Beneath a
Supersonic Turbulent Boundary Layer Perturbed by Steps and Shock Waves,”” NASA CR-486, 1966.
[46] Dolling, D. S., and Smith, D. R., ““Unsteady Shock-Induced Turbulent Separation in Mach 5 Cylinder Interactions,”” AIAA
Paper 88-0305, January 1988.
[47] Unalmis, O. H., and Dolling, D. S., ““Decay of Wall Pressure Field and Structure of a Mach 5 Adiabatic Turbulent Boundary
Layer,”” AIAA Paper 94-2363, June 1994.
[48] Gibson, B., and Dolling, D. S., ““Wall Pressure Fluctuations Near Separation in a Mach 5 Sharp Fin-Induced Turbulent
Interaction,”” AIAA Paper 91-0646, January 1991.
[49] Thomas, F. O., Putnam, C. M., and Chu, H. C., ““On the Mechanism of Unsteady Shock Oscillation in Shock
Wave/Turbulent Boundary Layer Interactions,”” Experiments in Fluids, Vol. 18, No. 1-2, 1994, pp. 69-81.
[50] Laganelli, A. L., Martelucci, A., and Shaw, L. L., ““Wall Pressure Fluctuations in Attached Boundary-Layer Flow,”” AIAA
Journal, Vol. 21, No. 4, 1983, 495-502.
[51] Sun, C-C., and Childs, M. E., ““A Modified Wall Wake Velocity Profile for Turbulent Compressible Boundary Layers,””
Journal of Aircraft, Vol. 10, No. 6, 1973, pp. 381-383.
[52] White, F. M., ““Viscous Fluid Flow,”” 2nd Ed., McGraw-Hill Inc., 1991, pp. 415-417.
[53] Clauser, F. H., ““The Turbulent Boundary Layer,”” Advances in Applied Mechanics, Vol. 4, 1956, pp. 1-51.
[54] Casper, K. M., Beresh, S. J., Henfling, J. F., Spillers, R. W., and Schneider, S. P., ““Pressure Fluctuations in Laminar,
Transitional, and Turbulent Hypersonic Boundary Layers,”” AIAA Paper 2009-4054, June 2009.
[55] Fujii, K., ““Experiment of the Two-Dimensional Roughness Effect on Hypersonic Boundary-Layer Transition,”” Journal of
Spacecraft and Rockets, Vol. 43, No. 4, 2006, pp. 731-738.
[56] Estorf, M., Radespiel, R., Schneider, S. P., Johnson, H. B., and Hein, S., ““Surface-Pressure Measurements of Second-Mode
Instability in Quiet Hypersonic Flow,”” AIAA Paper 2008-1153, January 2008.
[57] Zuckerwar, A. J., Herring, G. C., and Elbing, B. R., ““Calibration of the Pressure Sensitivity of Microphones by a Free-Field
Method at Frequencies up to 80 kHz,”” Journal of the Acoustical Society of America, Vol. 119, No. 1, 2006, pp. 320-329.
[58] Bradshaw, P., ““’’Inactive’’ Motion and Pressure Fluctuations in Turbulent Boundary Layers,”” Journal of Fluid Mechanics,
Vol. 30, No. 2, 1967, pp. 241-258.
[59] Panton, R. L., and Linebarger, J. H., ““Wall Pressure Spectra Calculations for Equilibrium Boundary Layers,”” Journal of
Fluid Mechanics, Vol. 65, No. 2, 1974, pp. 261-287.
[60] Blake, W. K., Mechanics of Flow-Induced Sound and Vibration, Academic Press, New York, 1986, pp. 497-594.
[61] Tsuji, Y., Fransson, J. H. M., Alfredsson, P. H., and Johansson, A. V., ““Pressure Statistics and their Scaling in High-
Reynolds-Number Turbulent Boundary Layers,”” Journal of Fluid Mechanics, Vol. 585, 2007, pp. 1-40.
[62] Hodgson, T. H., ““On the Dipole Radiation from a Rigid and Plane Surface,”” Proceedings of the Purdue Noise Control
Conference, 1971.
[63] Panton, R. L., Goldman, A. L., Lowery, R. L., and Reischman, M. M., ““Low-Frequency Pressure Fluctuations in
Axisymmetric Turbulent Boundary Layers,”” Journal of Fluid Mechanics, Vol. 97, Part 2, 1980, pp. 299-319.
[64] Klewicki, J. C., Priyadarshana, P. J. A., and Metzger, M. M., ““Statistical Structure of the Fluctuating Wall Pressure and its
In-plane Gradients at High Reynolds Number,”” Journal of Fluid Mechanics, Vol. 609, 2008, pp. 195-220.
[65] Simpson, R. L., Ghodbane, M., and McGrath, B. E., ““Surface Pressure Fluctuations in a Separating Turbulent Boundary
Layer,”” Journal of Fluid Mechanics, Vol. 177, 1987, pp. 167-186.
[66] Monin, A. S., and Yaglom, A. M., Statistical Fluid Mechanics, Vol. 2, MIT Press, 1975, pp. 375-377.
[67] Efimtsov, B. M., ““Similarity Criteria for the Spectra of Wall Pressure Fluctuations in a Turbulent Boundary Layer,”” Soviet
Physics –– Acoustics, Vol. 30, No. 1, 1984, pp. 33-35.
[68] Goody, M., ““Empirical Spectral Model of Surface Pressure Fluctuations,”” AIAA Journal, Vol. 42, No. 9, 2004, pp. 1788-
1794.
[69] Hwang, Y. F., Bonness, W. K., and Hambric, S. A., ““Comparison of Semi-Empirical Models for Turbulent Boundary Layer
Wall Pressure Spectra,”” Journal of Sound and Vibration, Vol. 319, No. 1-2, 2009, pp. 199-217.

-19-
American Institute of Aeronautics and Astronautics

You might also like