The GOES time code service, 1974-2004: a retrospective

10. Orbit Prediction Software

Orbit prediction software was always an essential element of the GOES time code service. The GOES satellite orbits were not stationary but changed with time as they were perturbed by several effects, the major ones being solar and lunar gravitational attractions, solar radiation pressure, and the inhomogeneous gravitational field of the earth. These perturbations can be modeled reasonably well, but the accuracy of the orbit predictions deteriorates with time due to residual imperfections in the modeling. High quality orbit prediction software that reduced these residual imperfections to a minimum was necessary for the service to meet its [+ or -]100 [micro]s timing specification, particularly because it was sometimes necessary to use orbital elements that were more than one month old.

[FIGURE 11 OMITTED]

The orbit prediction program originally used by the GOES service was a modified version of NASA's Goddard Trajectory Determination System (GTDS). This software consisted of about 40 000 lines of Fortran code [41], and was installed on a large Control Data Corporation 6600 mainframe computer in Boulder [29]. It was later installed on a similar computer at NOAA's facility in Suitland, Maryland, and if the Boulder computer was unavailable, the Suitland computer was accessed from Boulder. The input data for the software consisted of the six Keplerian orbital elements and their epoch [32]. The position estimates for the 240 h following the epoch were originally output on punched cards and brought to another computer where the cards were read into a file and then uploaded to the CDA, a slow process that new technology soon eliminated. The orbital elements were sent by NOAA via teletype until the mid-1990s, and in later years by email:

In the early 1990s, it was announced that both of the mainframe computers that ran GTDS were to be shut down, and NIST had to find new orbital prediction software. As part of a cooperative research program with the National Physical Laboratory in India (NPLI), NIST was given source code for the orbital prediction program used by NPLI's Indian National Satellite (INSAT) time service. Unlike GTDS, which was capable of predicting all types of orbits, the NPLI software was streamlined to work with geostationary orbits only. This source code was modified for the GOES data format and a personal computer software application called GOESTRAK was created, replacing GTDS in late 1992. GOESTRAK used a much smaller, more efficient algorithm than GTDS that relied heavily on iteration and proved to be equally accurate. NIST was beginning to receive orbital elements for the GOES satellites at longer intervals than before (sometimes more than 30 d elapsed between sets of elements), so GOESTRAK was designed to predict the satellite's position out to 50 d from the element's epoch. Even with 50 d old elements, the service could meet its [+ or -]100 [micro]s timing specification. Details of the orbital prediction algorithms were later published by NPLI [42].