Hopefully I can shed a little light on this.  The LDV techniques are
very fast for a single point, and sensitive only in the direction
perpendicular to the beam; this can be compensated by knowing the
incidence angle and correcting for the cosine error.  Basically you're
shining a single laser spot at the part, with two frequencies, and
looking at the shift in frequency as the reflection moves towards or
away from the detector. They are very sensitive, claiming motion
resolution to a few pm (1/100 the diameter of an atom!) though the
definition of how such motion is verified is not given in the
literature.  When this point is scanned across a test piece (about 1mm
smallest spot size), one can build up an out-of-plane motion plot of the
surface, though since each point of data is collected at a different
time, the phase relationship between the points is lost.  The systems
are best suited for membrane motion or single cantilever deflections,
which are well-constrained and primarily out of plane.  They are not
sensitive to in-plane motion at all, have a cosine decrease in sensitivy
from motion normal to the beam to perpendicular, and have problems on
diffuse surfaces.  Also, the measurement mode needs to be switched
depending on the frequency range of interest.  A part must be moving for
any information to be gathered, and they do not calculate static surface
parameters such as roughness, curvature, relative heights or spacings.
Recently they've been combined with bright-field strobing, which allows
in-plane only characterization to something like 30nm resolution (can
see features down to about 0.5mm), dependent on the optics being used.
The problem is that if a part moves both laterally and vertically (such
as a resonator that tilts as it translates) there is no measurement mode
to capture this.  For out-of-plane, you can calculate easily velocity
and acceleration as well as motion from the data given.



Optical profilers are full-field devices, and gather information from
all points in the field of view of the objective simulateneously.  This
maintains the phase relationship between points, and also means that you
gather X,Y,Z data about every point, so you see both in-plane,
out-of-plane, or a combination of the two.  They use the 3D data to
segment the part and you can do calculations such as radius of
curvature, relative tilts, lateral motions, vertical motions, roughness,
linewidth all with the same measurement.  They are not real-time,
however, requiring the user to step through frequency and/or phase to
build up a complete picture of the device across its entire range .
Typical measurements take between 1 and 5 seconds per frequency, but you
are getting a few hundred thousand surface data points in that time.
You can still do resonance frequency, look at transient motion,
switching time, settling time, but also have the basic 3D static data to
work with as well.  Vertical resolution is sub-nm, and lateral
resolution is about 0.5microns for seeing static features, with lateral
motion resolution on the order of 20nm (both lateral figures are
dependent on the objective, with higher mags giving the best resolution.
These systems have been around for about 20 years, so the software is
pretty advanced for surface characterization, both for static and
dynamic properties.



A lot of information, but hopefully helps a little.  Each system has
advantages, though the more flexible is the optical profiler in that it
can do static and dynamic characterization.  If you are going to do very
high volume, very long frequency sweeps, and are mainly interested in
out-of-plane, the LDV systems are going to be faster at that.



-Erik