[FOSS-GPS] Proper formula to calculate R95 accuracy circle from RTKLIB

Jeff Hayward jeffhayward at gmail.com
Fri Nov 21 21:27:53 PST 2014


Hello FOSS-GPS,

I need some help computing error (accuracy) circles from RTKLIB.

is R95 = 1.74 * sqrt(sdn**2 + sde**2) the correct formula for a 95%
confidence interval accuracy metric?

Background (sorry,long):

I am using RTKLIB to post-process track logs from a backpack-mounted
rover against CORS/UNAVCO reference station data and would like to
compute the error circle for each position in the result.

My problem is that the error circles I'm computing look way too good
to be true.

We get successful positions, and they check well against both the
onboard engine (u-blox NEO-7P) and other knowledge of the track.
FYI, our application is mapping long-distance trails.  Volunteers carry
custom loggers and make trail notes as they hike.  We record raw
pseudorange and carrier information as well as the real-time points
the onboard engine computes.

I'm by no means a PNT expert, so please help me understand how most
GPS device are calculating the error circle, or "accuracy" metric.

I've been taking the 'sdn' and 'sde' values reported in the .pos file
and using e.g.

    RMSE = sqrt(sdn**2 + sde**2)

as the basic 2-dimensional metric and then scaling it by e.g

   R95 = 1.74 * RMSE

To get an R95 circle radius.  But the resulting values are so good
that we don't really believe them.  We regularly see positions < 10 cm
R95 which is not usual given a comparison to survey-grade receivers
such as a Trimble GeoHex + Tornado antenna we have used in the past.
Also the corresponding point from the NEO onboard engine will
typically be in the 2+ meter range.

We're post processing L1 carrier phase with rnx2rtkp 'kinematic',
using IGS precise ephemeris and clock files and the reference station
files from UNAVCO.  Our baselines are typically 3-30 km.

Interestingly we also get very small error numbers when we process
using single-receiver ppp-kinematic and the IGS precise ephemeris and
SBAS ionospheric.

But the numbers as I'm calculating them seem just too good to be true.

Thanks for any help.
--
Jeff Hayward
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/foss-gps/attachments/20141121/727b6b41/attachment.html>


More information about the FOSS-GPS mailing list