Practical limits of SDP4

Bruno Tilgner (
Wed, 29 Jul 1998 17:36:10 -0400

On 28.07. Jason Hatton wrote:

> I was wondering what the practical limits of the prediction
> algorithms used in tracking software are for satellites in
> "deep space" orbits. I know that SDP4 works quite well for
> satellites in geosynchronous orbits (ie. 36,000km alt)
> & in geosync. transfer (ie 600km x 36,000km)......

I should like to add a word of caution as regards the accuracy
of SDP4 for geostationary orbits. Last February, an observer
in southern Switzerland and myself have compared during 18
consecutive nights the positions of the then 7 ASTRA satellites
as observed with a CCD camera through an 8" telescope with positions
calculated with SDP4 from NORAD TLEs. The accuracy of the
observations was well below 1 arcmin.

The SDP4 model not only gave far too high longitudinal drift rates
for all satellites, it also failed completely to reproduce their
daily motion which we have observed for about 12 hours every night.
I have not been able to determine whether this failure is due to
inaccurate orbital elements or to a flaw in the SDP4 model, not
necessarily the underlying algorithms but possibly the way they have
been implemented in FORTRAN and subsequently in PASCAL code. (Earlier
last year a rather trivial programming error was discovered by
Rainer Kracht, I believe).

The TLEs for at least one of the ASTRA satellites (1F) was grossly
wrong for several months. The position resulting from these TLEs
in the end was some 20 degrees farther east than the true position.
More subtle inaccuracies may have crept into the other TLEs.
Unfortunately, SES (Societe Europeenne des Satellites), the owner
and operator of the ASTRA satellites, refused to communicate the exact
orbital position of the satellites, presumably for fear of an attack
by high-tech hackers.

It may be worth noting that SDP4 does not give noticeably different
results from SGP4 for periods in the order of several days. Luni-
solar effects only come into effect at geostationary altitudes
over longer time spans. In this particular case they were completely
masked by position errors caused either by inaccurate TLEs or by
flaws in SDP4.

Bruno Tilgner
Paris, France