Synthetic Seismogram normalizations

I have received queries about constants in the synthetic seismogram code anirec.f

In order to avoid round-off error and numerical overflow in numerical computations, it is conventional to nondimensionalize physical units in synthetic seismogram calculations. My codes use the convetion that I inherited from Freeman Gilbert, my dissertation advisor, and the UCSD Free Oscillation crowd. The rules for normalizing Mass, Length and Time are derived from three constraints.

density is normalized by rho-bar = average Earth density = 5515 kg/m^3

length is normalized by the mean Earth radius R=6.371*10^6 m

The last constraint is that pi*G=1, where pi=3.14159265358979d0 and G=6.6723*10^{-11} m^3/(kg-s^2) is the Gravitational constant.

From the public free-oscillation code MINEOS

c the model has been normalised such that a density of 5515 mg/m**3 = 1 ;
c pi*g=1 where g is the grav constant (6.6723e-11) and the radius of the
c earth (rn=6371000 m) is 1. these normalizations result in
c acceleration normalisation = pi*g*rhobar*rn = 7.365081 m/s^2
c velocity normalisation = rn*sqrt(pi*g*rhobar)= vn = 6850.0396 m/s
c frequency normalization = vn/rn = sqrt(pi*g*rhobar) = 1.075190645e-3 s^1


This last constraint leads to a normalization constant for frequency of REN=1.075190645d-3. I have found a slightly different value for G quoted on the web, which is understandable considering that the seismological constants were set when G was not known as precisely as now. There can be danger in mistyping the numbers. One colleague of mine lost days and much sleep when the rho-bar constant was mistyped 5517 kg/m^3 and a 10^{-4} error crept into calculations that should have been good to double precision. Believe me, no one was lauging when the error was discovered!