NOTE:  y is UP in the PIV datasets (elsewhere z is typically UP and y is SPANWISE)

Units of x and y in mm (with x=0 mm at centroid, y=0 mm at bottom wall)
  (height of model y = 6 inches = 152.4 mm)

Units of velocity in m/s
Units of TKE in (m/s)^2
Units of turbulent stresses in (m/s)^2

Note that the data in Centerline_FAITH_2Hz_4000samps_scalar were replaced on
12/9/2014, in order to revise the axis to have y=0 mm at the bottom wall (like
in the Centerline_FAITH_1kHz_3940samp_scalar data).
Also please read the following notes from the experimentalist:

For the 2 Hz data, the apparent data occurring inside of the body comes from 
correlation noise on the body that gets averaged with the flow data. A mask was
applied on the body, but that mask is based on the perspective of the camera 
which is necessarily distorted from the “ideal” cosine curve. The distortion 
could not be corrected because the model lies outside the calibration plane.

The two frequency designations were acquired about a year apart. The 2 Hz are 
from 11 Mpix cameras that operated at that speed. The experimentalists 
subsequently acquired a high-speed PIV system that uses 4 Mpix cameras up to 
730 Hz. They reduced the image size somewhat to get 1 kHz PIV frequency. The 
laser power for the two systems was quite different. The High speed PIV used 
a laser that gave about a 20 mJoule/pulse, whereas the low frequency system 
used a laser with ~400 mJoule/pulse. To gain efficiency of light scattering, 
the high speed cameras were moved upstream to increase the forward scatter 
angle. Therefore the hill shadowed a portion of the laser light sheet in the 
1 kHz data. This manifests itself as missing data near the surface, or 
"shadowing."

