Re: Matlab to Octave - precision and differential equations

[Sergei Steshenko]

--- On Mon, 10/4/10, Dotan Cohen <address@hidden> wrote:

> From: Dotan Cohen <address@hidden>
> Subject: Re: Matlab to Octave - precision and differential equations
> To: "Sergei Steshenko" <address@hidden>
> Cc: "fork" <address@hidden>, "Doug Stewart" <address@hidden>, address@hidden
> Date: Monday, October 4, 2010, 5:17 AM
> On Mon, Oct 4, 2010 at 14:05, Sergei
> Steshenko <address@hidden>
> wrote:
> > You rather probably need a different numerical
> method.
> >
> 
> Depends on whether he is doing research or development. For
> modeling
> gravity to determine a natural object's trajectory through
> the solar
> system, four or eight significant figures might do. For
> actually
> sending a probeto Neptune via a Jupiter gravity assist, he
> will need
> considerably more!
> 
> -- 
> Dotan Cohen
> 
> http://gibberish.co.il
> http://what-is-what.com
> 

Well, still debatable. I.e. the core issue can still be the numerical
method.

A known method is to take a differential equation with known analytical
solution, preferably not very "smooth" solution, and to solve it using
the numerical method under question.

Then to compare the solutions.

I think most actual orbits have been calculated using FORTRAN and
what is called nowadays "double float" (in "C" terms) type.

Regards,
  Sergei.