REYNOLDS STRESS STRUCTURE IN TURBULENT BOUNDARY LAYER 291
valid data because the data signal itself is usually used to develop the charge
on the capacitor. In measuring average values rather large time constants
were required to obtain stable averages during sampling measurements of
infrequently occurring events. When digital methods were used, the amount
of data required to obtain stable average values was approximately 5 % of
the data required using analog methods.
We shall describe only results obtained using digital data reduction
methods. All of the digital data reduction was done using prerecorded data
reproduced at oneeighth the real time recording rate. After the initial exploratory measurements had baen made and the experimental set up was
designed, the actual data collection and recording prooess required only two
weeks time. Approximately two years were spent writing programs and
using them to reduce the digitized data.
3.1. Conditionally Sampled Measurements of Reynolds Stress
The method of conditional sampling was first used by Kibens (1968; for a
summary, see Kovasznay et al., 1970) in the study of the motion and shape of
the turbulent bulges in the outer intermittent region of a turbulent boundary
layer. We have extended their sampling c o n q t s to aWv the extraction of
individual contributions to Reynolds stress from a ffiUy turbuknt signal. To
separate the Reynolds stress contributions from the background turbulence
signals one needs a criterion to dacide when the Reynolds strcss contribution
occurs. In our study of the flow field near the wall we were guided by the
results of Kim et d. (1968) and Corino and Bodkty { 1%9) which show that
before an eruption or bust occurs the fluid beneath the region of eruption
attains an unusually low streamwise velocity.
We used the simple criterion that when the streamwise velocity u, very
near the wall becomes low and decreasing with negative slope (i.e.,
du,/dr < 0) an eruption or burst is likely to occur. To test the validity of this
criterion we sampled the instantaneous record of the product UD obtained
from an x configuration hot wire anemometer probe for a set of nine different values of u, (- 2 4 < u,, < 24,) for both positive and negative slope.
Figure1 is a sketch showing the configuration of the hot wire probes and the
coordinates denoting the position of the hot wire probcs.
The basic scheme we used is outlined in the sketch of Fig. 2 in which a
single sample of the uu signal b acquired in a " window " centered about the
point at which the signal u, has attained, for exampk, a certain specified
level with positive slope. The sampling was performed by an IBM 360/67
computer which was programed usin8 FORTRAN to acquire samples from
all nine levels of u, with both positive and negative slopes in one pass
through the data recorded on digital magnetic tape. Thesamples for each of
valid data because the data signal itself is usually used to develop the charge
on the capacitor. In measuring average values rather large time constants
were required to obtain stable averages during sampling measurements of
infrequently occurring events. When digital methods were used, the amount
of data required to obtain stable average values was approximately 5 % of
the data required using analog methods.
We shall describe only results obtained using digital data reduction
methods. All of the digital data reduction was done using prerecorded data
reproduced at oneeighth the real time recording rate. After the initial exploratory measurements had baen made and the experimental set up was
designed, the actual data collection and recording prooess required only two
weeks time. Approximately two years were spent writing programs and
using them to reduce the digitized data.
3.1. Conditionally Sampled Measurements of Reynolds Stress
The method of conditional sampling was first used by Kibens (1968; for a
summary, see Kovasznay et al., 1970) in the study of the motion and shape of
the turbulent bulges in the outer intermittent region of a turbulent boundary
layer. We have extended their sampling c o n q t s to aWv the extraction of
individual contributions to Reynolds stress from a ffiUy turbuknt signal. To
separate the Reynolds stress contributions from the background turbulence
signals one needs a criterion to dacide when the Reynolds strcss contribution
occurs. In our study of the flow field near the wall we were guided by the
results of Kim et d. (1968) and Corino and Bodkty { 1%9) which show that
before an eruption or bust occurs the fluid beneath the region of eruption
attains an unusually low streamwise velocity.
We used the simple criterion that when the streamwise velocity u, very
near the wall becomes low and decreasing with negative slope (i.e.,
du,/dr < 0) an eruption or burst is likely to occur. To test the validity of this
criterion we sampled the instantaneous record of the product UD obtained
from an x configuration hot wire anemometer probe for a set of nine different values of u, (- 2 4 < u,, < 24,) for both positive and negative slope.
Figure1 is a sketch showing the configuration of the hot wire probes and the
coordinates denoting the position of the hot wire probcs.
The basic scheme we used is outlined in the sketch of Fig. 2 in which a
single sample of the uu signal b acquired in a " window " centered about the
point at which the signal u, has attained, for exampk, a certain specified
level with positive slope. The sampling was performed by an IBM 360/67
computer which was programed usin8 FORTRAN to acquire samples from
all nine levels of u, with both positive and negative slopes in one pass
through the data recorded on digital magnetic tape. Thesamples for each of