(Received 19 September 1966) This paper is a discussion of the resolution by a piezo-electric transducer of the local pressure fluctuations at the wall of a turbulent boundary layer. The first part applies to transducers whose sensing area is large in comparison with the boundary layer thickness, and the second part to transducers whose sensing area is either of the order of, or smaller than, the boundary layer thickness. For large transducers, the main result, derived for a sensing surface of arbitrary shape and with an arbitrary spatial sensitivity distribution, is that under circumstances which are perhaps unrealistically stringent, the attenuation of a given frequency component will increase as the cube of the transducer face linear dimension. In general, it should be expected to increase only as the square of the dimension. In the second part, the probable dependence of the resolution on frequency and on transducer size is indicated by using both recent experimental evidence and physical reasoning bearing on the nature of the pressure cross-spectral density. Both approaches suggest that in a co-ordinate system translated downstream at an appropriate velocity, as the streamwise wave number increases, the spectral density in wave number-frequency space is associated with proportionately increasing spanwise wave numbers and almost proportionately increasing frequencies. Experimental evidence suggests departure from this finding if the fixed axis longitudinal wave length is very large (of the order of the boundary layer thickness) in which case both the characteristic time in convected axes and the characteristic lateral coherence length appear to reach asymptotic finite values. For shorter waves, the strong relationship between the two space scales and the convective time scale necessarily implies that the resolution of the frequency spectral density by a finite transducer continues to deterioriate as the frequency increases. It is concluded that the attenuation predictions previously published by the author are not likely to be seriously in error. I.
INTRODUCTION
When a pressure transducer of non-zero sensing area is mounted as part of the solid boundary surface of a turbulent boundary layer it responds to some area integral of the instantaneous pressure found on the boundary. The manner in which the output signal thus fails to measure pressure at a point has been discussed by Corcos [I] (henceforth referred to as [I]), Gardner [2], White [3] and Willmarth and Roos [4]. In what follows we first consider again the case of large transducers by removing assumptions which are not strictly necessary and by focusing on a theoretical constraint which is provided by the equation of motion and which was overlooked in [I]. The notation is that in [I]. For two identical linear transducers with a perfect time response, mounted as part of (but uncoupled from) a solid surface along which the turbulent pressure field is statistically stationary and homogeneous, it has been shown (see, e.g., [I]) that the measured crossspectral density is given in terms of the actual cross-spectral density by
60
G. M. CORCOS
where c is the distance vector between the two transducer Thus, the measured frequency spectral density is Q&J)
= rm(o, W) = ,I
r(e,
positions,
and w is the frequency.
CO)O(r) dA(s),
(a)
where O(E) = j- K(s)K(s+ m
E) dA(s).
Here K(s) is the unit response of the transducer, i.e. the contribution to the transducer output signal due to a unit pressure at point s on its face; i.e. the transducer located at x measures a pressure p, given by P,(x,
t) = j” P(S, 4 K(s - 4 dA(s).
Hence O(E) is a function wholly defined by the transducer shape and characteristics. E is the two-dimensional vector (E%,E~)defining the co-ordinates on the transducer face, c = (4,~) is the separation between two identical transducers used for the measurements, w is the frequency and i- 00 r&CO)
= ,‘,
W,
I --co
7) exp [ - i(+]
d7,
where R(<,T) is the space-time covariance or correlation of the pressure at two points separated by the distance vector < and at two times separated by the delay 7. 2. ASYMPTOTIC DEPENDENCE SPECTRAL DENSITY
OF THE MEASURED FREQUENCY ON TRANSDUCER SIZE
The behavior of transducers of large sensing face dimensions is of interest in connection with sonar design. It was briefly considered in [I]. For transducers of given shape and characteristics, with a typical face dimension L, of L, is a complex we have O= @(e/L), whereas r(<, w), which is evidently independent function whose modulus at a given w is a maximum at < = o and which decreases to zero for any w. as \
‘1
l
c3 /
,--4
&
i
I.
\--
?
‘L_+’
Figure
1’
g
The area of integration.
We consider the open set ~2 of points in the plane whose boundary is a perimeter V described by the locus of the end of the distance vector between the origin of two identical transducers of identical orientation if one is held fixed and the other is translated to assume all possible contacts with the first (i.e. if the area of the transducer defines the set of points a, then the set &‘=g--a). The characteristic length for the perimeter $5’is
TURBULENT
PRESSURES
ON A BOUNDARY
61
LAYER
called L. We assume that 0 is a continuous function of (e/L), non-stationary at the origin, with continuous first derivatives on & and identically zero on and outside V. We also assume that the transducer is sensitive to pressure rather than to force, so that 0 is of the form 0 = if(E/L), where A is the sensing surface area of order L2 andf(s/L) Then we expand 0 in a partial series