I’ve read all topics related to my problem but unfortunately no progress. That is why I decided to open the new one.
Here is a task: I consider a particular GEO. For this target, I have the “True” (known precisely from a source different from TLE) position of GEO and TLE. I would like to compare the positions generated by SGP4 with the “True” one. The “True” vector is given in EME2000. Below is part of my code to get PV from TLE in EME2000 frame. The difference is plotted in the attached figure. I think the difference is too large. What did I do wrong?
I don’t see any obvious error in your code.
I would say the TLE is wrong. I remember a presentation by T.S. Kelso in 2011 where he showed an error for a satellite in GTO that was … 60000km, and he insisted on the fact the plot was in kilometers, not meters, that was 1.4 times the geosynchronous semi major axis. Of course, it was an extreme case, with valid reasons explaining the discrepancy (for the record it was due to a LEOP maneuver not considered by NORAD that just extrapolated previous observations instead of using an updated orbit, despite the operator did provide the information to them directly).
Could you check with other TLEs at other dates?
Thanks for your prompt reply and information regarding TLE. I was aware of the situation with TLE but did not that it is that bad. I repeated my comparison with another (new) set of TLE and the result become better. This one more time confirm that if using TLE stay with SGP4. I was wondering if I may use two different sets of information when computing Near-miss events between two GEO. For one I know “precise” PV at a particular moment and for the other I was going to use TLE for the same (close) epoch and propagate them into the future for 5-7 days. Seems it is not going to provide me with sufficient precision. If I use TLE for both GEO then Near-miss event prediction is +/- 2 km in distance (between GEOs) and +/- 1 sec in time (estimated using observations of predicted events).