Re: {MPML} 2002 JR100: Find_Orb bug, workaround

Bill J Gray May 5, 2010

Hello all,

Paulo Holvorcem recently mentioned on this list that he'd had a
problem with a bad orbit solution from Find_Orb. That problem
turns out to be a bug. For the nonce, I'd recommend people edit
the file 'environ.dat' and add this line:

COWELL=1

and the problem will go away. The only effects visible to you, the
user, should be that orbits converge the way they're supposed to.

It appears to affect only longer arcs. It comes about because
Find_Orb supports two different integration schemes, that of Encke
(the default) and Cowell. Something's amiss with the Encke code
(still haven't figured out what).

In theory, the Encke version ought to be faster. But in most
cases, it doesn't seem to matter very much, so I don't think anyone
is apt to lose much by using COWELL=1.

Unfortunately, even if I _do_ figure out the problem right away,
it may be a bit before I can post an update. Some discussions with
Sonia Keys at MPC, and some of the information in Milani and Gronchi's
very interesting book _ Theory of Orbit Determination_, have gotten me
working on adding statistical ranging to Find_Orb. I'd like to get
that finished before I post an update.

If you run into questions about this, please e-mail me directly,
or post on the Find_Orb list, or both.

-- Bill