Re: [find_orb] covar.txt not getting written out with the Apr 19 2012 DEV version?

Bill J Gray May 20, 2012

Hi Peter,

It took me a little while to figure out a fix for this. I've just
uploaded a new development version to

http://www.projectpluto.com/pluto/devel/find_orb.htm

in which sigmas are re-enabled.

HOWEVER, you may see trouble if you set an epoch that isn't very
close to the observed arc, especially if there are close planetary
encounters between the epoch and the observed arc. (Which is why I
disabled covariance/sigma computation in the recent version.)

For an example, try 2012 DA14, the object slated to make a
close approach on 15.8 February 2013. You can solve for its orbit
in Find_Orb without much trouble (load data, turn on all
perturbers, click Auto-Solve). Set an epoch before about 13 February
2013 and do a full step, and all will work correctly.

But set an epoch of, say, 17 February 2013, after its flyby, and
do some full steps, and the orbit diverges. Essentially, Find_Orb's
efforts to describe an uncertainty "ellipsoid" don't match the actual
shape of the uncertainty area, which is really more of a "banana-noid"
or some truly bizarre corkscrew shape.

You may never run into this issue (unless you're examining the future
behavior of objects such as 2012 DA14). If you do see diverging orbit
solutions in Find_Orb, though, there's a fix. With the new
version, you can edit 'environ.dat' and add a line such as

NO_SIGMAS=20

to tell Find_Orb: "Don't attempt to compute covariance data for
epochs that are more than 20 days from the observed arc." If you set
such an epoch and then did a full step, 'covar.txt' wouldn't be
updated and the uncertainty parameter U wouldn't be shown in the
orbital elements.

For 2012 DA14, one could actually set that to closer to 300 days.
I'd really prefer either to solve the numerical instability issue. But
at least for the nonce, this will allow one to get covariance data
almost all the time, but to disable it in cases where it causes trouble.

Hope this all makes sense... please let me know if it doesn't!

-- Bill