Question about the age of available TLEs

From: Gavin Eadie via Seesat-l <>
Date: Fri, 9 Feb 2018 20:02:15 -0500
I retest my TLE processing library when I introduce any improvements by comparing my output with output from other sources.  Heavens-Above has fairly high precision output and I often use that for comparison.  I noticed, a few minutes ago, a discrepancy.  I’m not suggesting this is a new matter, nor that any errors are involved on anyone’s part (except maybe mine!), nor that this is of any serious concern, but I am curious.

At about 2018-02-08 22:55 UTC I ran my tests using the latest Celestrak ISS TLE and noticed a small difference from H-A results for the next visible pass.  The difference was minor, maybe about ten seconds in time; but my library is accurate .. given the same TLE input, I should get the same output as anyone using Vallado’s [AIAA 2006-6753] algorithms, so I was concerned that I had introduced an error into my code.

The short version of the resolution is that the Celestrak TLE epoch I used differed from the H-A TLE epoch by about one and a half days.  Still being curious, I did some more digging and found that H-A was using the current "Coasting ARC #2" elements from NASA’s "Realtime Data" web page (TLE epoch on Sunday).

So my algorithmic concern is soothed, but I remain vaguely curious.  Here are the concurrent TLEs from the three/four sources (concurrent meaning captured within ~5 mins):


1 25544U 98067A   18039.95265046  .00001678  00000-0  32659-4 0  9999
2 25544  51.6426 297.9871 0003401  86.7895 100.1959 15.54072469 98577


1 25544U 98067A   18040.57458473 +.00001514 +00000-0 +30164-4 0  9995
2 25544 051.6426 294.8869 0003344 087.4701 341.3335 15.54074418098674

NASA Realtime ARC #1

1 25544U 98067A   18040.56232320  .00016717  00000-0  10270-3 0  9003
2 25544  51.6421 294.9562 0003081  79.1822 280.9678 15.54068781 18665

Heavens-Above (same as NASA Realtime ARC #2)

1 25544U 98067A   18041.52688660  .00016717  00000-0  10270-3 0  9013
2 25544  51.6419 290.1489 0003003  84.4747 275.6748 15.54054067 18815

I know there’s nothing untoward about future TLE epochs.  I’m also not surprised Celestrak and Space-Track are different given Celestrak probably samples S-T less frequently than S-T data changes.  I also suspect any differences in computed positions and times are probably less than SGP4 fidelity anyway, and so my little trip through this story may be of no practical significance but: 

Question:  Assuming the non-trivial variation in TLE epochs for the same satellite from different sources was a surprise to me, what collective wisdom have I missed?  Are future TLE epochs somehow "better"?  I’ve never seen any conversation about this .. maybe it took place decades ago?!
Seesat-l mailing list
Received on Fri Feb 09 2018 - 19:03:13 UTC

This archive was generated by hypermail 2.3.0 : Sat Feb 10 2018 - 01:03:13 UTC