At 06:59 AM 16/06/2003 +0200, you wrote:
Hank de Wit wrote:
However, the method of simple differences introduce a 12 hour phase error
so we would be better off producing the differential dEOT/dt. As the
fourier approximation is linear this can be done with high school
calculus. I've included the differential function below. Luckily this
produces the same numerical result (to two dec. places) as before except
the date is now 23.0UTC Dec (as expected from the phase argument).
Beware! The derivative of an approximating function need not be the
approximation of the derivative of the real function.
Very true. However, in this case, because the EOT is a "well behaved"
function and well described by a truncated fourier approximation, I think
it is valid to use the derivative of the approximating function. Besides
there is no "real" EOT function, all existing algorithms are approximations
of varying accuracy.
Besides, we must take into account that all algorithms to calculate the
EoT aren't very robust and are only accurate for
a more or less narrow span of time. No algorithm would be able to
calculate the EoT on the day when Ramses II was
born, for instance.
True again. This approximation is only really valid for 2002. I wasn't
really trying to argue this was a high precision method, just adequate for
many purposes, and simple to calculate.
And finally, if John or somebody else wants to work in the range of 10 sec
they'll have then to take into account other
difficult to calculate factors like the atmospheric reffraction (the
formulae we know are all rough approximations).
Yep.
Best regards,
Anselmo Perez Serrada
Cheers
Hank
-
