 need to interpolate in depth to regular depth intervals,
 nearest neighbour suffices.


 over how many points to smooth ?

lamda_min = c_min/f_max
lamda_max = c_max/f_min


for cmin=1500 and fmax 50  ->  lmin=30m     -> 300pts
for cmin=3000 and fmax 10  ->  lmin=300m    -> 3000pts


partially integerated velxp  is pretty damn close to p vrms.


How about now Nmo velocities ?



method -> equivmed -> S+M

gives closer relationship to observed
than just smoothing + blocking  1/vp

eigen value average versus full matrix average 
(are identical if eigenvectors of the averaged quantities are identical)
or Vt * U = 1 )


- need to have shear sonic  or need to assume vs  (generate empirically)


+ include dip information correctly into average.


- assume initially iso  ->  end up with ti  slightly more complex

+- francis approxmiation works in vicinity of the well.




NEED to show that amplitudes are modelled correctly:

solid and FD
