[Qgis-user] wishing for accurate lattitude/longitude from a cell, phone

[Garth Fletcher]

I have added some data to my earlier report, both from my Garmin eTrex
and from a Bad Elf Surveyor.

As described earlier I established an accurately measured test location
(± 2.4 cm) and taken 17 readings with my Garmin eTrex 20, distributed
over a 10 hour period. The eTrex was receiving both US GPS and Russian
GLONASS satellites, plus WAAS (as indicated by a "D" in the Garmin's
satellite signal bars).

These points were plotted on a 1 meter grid with a link provided in my
earlier report to the PDF graph.

That graph
   <http://www.Mason-NH.org/Specials/GARMIN_TESTS.pdf>  (56 KB).
has now been updated with more Garmin eTrex 20 data plus some data using
a Bad Elf Surveyor. On the updated graph the original 17 points are
plotted as blue unlabeled dots.


I added another 12 pairs from the Garmin eTrex, taken at the same
location at variable intervals throughout a 10 hour period on May 27. The
first reading in each pair was taken (Mark Waypoint) and stored,
followed by a second reading in which I told the Garmin to "average
waypoint" before storing. This averaging takes from 4 to 10 minutes
until a bar reaches "100% confidence", whatever that means.
Note that this is not Garmin's recommended use - which is to average
waypoints with at least 90 minutes between.  However my immediate
averaging should filter out short term "noise".

These points are plotted as green (immediate) and red (averaged) dots,
labeled x and xA, e.g., 1 and 1A, for the immediate and averaged points,
respectively.  I also added thin magenta lines linking the two dots in
each pair.

That data indicates a ±7 meter "wander", even in the averaged data.


I also set a Bad Elf Surveyor at the same location and recorded data
over a 2 hour period into a RINEX file. The estimated position returned
by CSRS-PPP post processing (Rapid) is plotted as a yellow star. CSRS
also reported a distressingly large error ellipse: 10.8 x 8.7 meters!

While the Surveyor is recording via its Bad Elf app one can also display
a "Deviation Plot" which shows the wandering of the measured location,
centered over its recent average (past 15 minutes).  This "wander" was
observed to drift slowly within a 20' (6m) radius circle; i.e., in any
15 minute period the maximum drift seen was ±20' (up to 40'), but could
have been larger over longer intervals.

While the Surveyor is described as being compatible with US GPS and
Russian GLONASS, it can use only one constellation at a time.
Consequently the Surveyor tracks many fewer satellites than the eTrex
which tracks both GPS and GLONASS.  This further reduces its performance
in less than the optimal "360° open sky" environments.

The Surveyor, like the eTrex, is also a single-frequency receiver, thus
its RINEX recordings cannot be post-processed by USGS OPUS.  CSRS-PPP
will process single-frequency data (includingg GPS + GLONASS), but is
constrained by the limitations of single-frequency data and wandering
data.


My conclusions:
under realistic field conditions (not all satellites visible) neither
   the Garmin eTrex nor the Bad Elf Surveyor can be relied upon for
   mapping-grade accuracy.
the Surveyor's RINEX capability is of very limited use due to its
   single constellation and single-frequency limitations.
the slow "wander" in single-frequency GPS location suggests that
   averaging, even if done with several widely spaced observations, is
   unlikely to improve accuracy.  It would be necessary to average many
   dozens of readings evenly distributed throughout a satellite orbital
   period (~12 hours) in order to get a reasonable sampling capable of
   reducing uncertainty.
-- 
Garth Fletcher