How important is a Precise Ephemeris?

This question has bugged me somewhat in the past several months..Many say it's not something to worry about and won't affect your data all that much...unless your are doing long baselines..

The past several nights, I have been re-proccessing a hundred or more baselines using the precise ephemeris...It does make a difference, but I'll let you see the results and draw your own inferences from the data...

Without fail, EVERY baseline, no matter how short (I even did several that were 1000' or so)had a significant drop in the RMS, a drop in the RefVar and an increase in the Ratio...I know that these are just statistics, buy hey, it's all we got to work with using GPS..

A couple of samples;
First.I did no data cutting, mask raising or any other manipulation of the data..The first will be the non-precise and the second the precise..Precisions and distances will all be in US Survey feet (I know, I know, but I'm not doing the metric thingy)...

Two-Porter
Ratio RevFar RMS HorzAc VertAc Dist
3.2 2.033 0.011 0.009 0.024 9.6nm
4.4 1.383 0.009 0.007 0.017
Porter-383A
26.6 1.312 0.004 0.003 0.005 1.5nm
28.9 1.260 0.004 0.002 0.004
383A-Elgin
2.1 5.645 0.009 0.014 0.016 4.5nm
3.1 4.270 0.007 0.013 0.017
Porter-Tahoe
6.7 7.027 0.008 0.007 0.016 8.1nm
11.6 3.794 0.006 0.005 0.012
Porter-Apache
3.7 2.849 0.006 0.003 0.011 4.0nm
5.0 2.066 0.005 0.003 0.010
383A-383B
14.8 1.237 0.004 0.002 0.005 0.3nm
19.8 0.973 0.003 0.002 0.005

And the list goes on and on.....I'm satisfied at this point that the precise makes real good statistics, but what does it do to the point-point...

I won't list another long line of distances, but will relate what showed up in the data.. I compared the computed distance with non-precise versus precise...

0-2 miles (difference at the tens of thousandth's of a foot, and random pattern, long and short)No worries mate!

2-5 miles (up to 0.005' near 5 miles). All the precise distances were shorter (I think this is just a coincidence)Again..Don't think I'd fret to much..

5-10 mile (up to 0.01 beginning right near the 6 mile mark and staying fairly steady up to 10)... A hundreth...GOOD GOD MAN, the world is coming to and end...Seriously, it has begun to show up on each line...

Can we deduce that the systematic error induced by the broadcast ephemeris can affect your distance by 0.01? I'd say yes, and also say, I wouldn't worry much about it in the normal routine of the day...but since our project is meant to reduce any error that we can get rid of, then it appears to be measurable.....

Will the better stats affect the least squares significantly? Don't know the answer to this one yet, but suspect that some points will fudge, but probably not enough to be measurable.

Also followed the time line and the SV tracking pretty close when processing this data..Bad SV's looked even worse in the precise...Which would mean easier to spot and get rid off when processing...

My conclusions....
The precise would be very helpful to use, but they need to be posted quicker for the average working surveyor. Most projects cannot wait 14 days for the precise to be put on the web...
The precise definately affects the vertical more than the horizontal..I suspect that much better coorelation will be possible when geoid modeling when using a precise eph to process baselines...

And finally, a question for the NGS...Is there a factor of accuracy between the ultra-fast-precise eph's as posted...Maybe we only need the fast to accomplish our goals..Dont' have a clue about this, but they show up a day behind the data instead of 14 days...

And really, finally, I also have some very long lines done the same way as the above...I was wondering if the error keeps ramping up or dies at so

My WAG (not SWAG) would be a parabola...but the baseline lengths to neutralize the orbital differences might be considerably greater than anything we have thought about trying.

Jimbo

Merry Christmas