Error on successive days

Does anyone have any idea about the likely cause of this please. The
screenshot shows two GPX files in Google Earth of a 2.6 mile walk. The first,
recorded starting mid-afternoon on Monday 12th April 2010, looks pretty
accurate in relation to GE's road layer. (The short excursions about three
quarters of the way up was genuine.) But the second, on Tuesday, is
significantly displaced.

This was a steep road climbing about 1000 ft from the town to the top. It had
no obstructions by trees.

Both were recorded on my smartphone, an HTC Touch Pro 2 and later copied to
my PC using Memory-Map.

http://dl.dropbox.com/u/4019461/20100412%2B13-GPSerror.jpg

--
Terry, East Grinstead, UK

From Google Maps, using these coordinates:

37.538759, -4.30913

looks like roughly a 50 m East bias.

Like having a time error of about 135 ms at that latitude.

Are you using QuickGPS for ephemeris?

http://community.htc.com/na/htc-forums/windows-phone/f/23/p/460/2619.aspx

http://www.htc.com/europe/faqs.aspx?p_id=291&cat=266&id=111054

HTC offers some sort of patch here, but doesn't really explain what
it's supposed to do (i.e., if it ain't broke, don't fix it):

http://www.htc.com/europe/SupportDownload.aspx?p_id=250&cat=0&dl_id=624

Thanks Ed, appreciate the reply.

Yes, I do use QuickGPS, but usually only if I remember before starting my
application (Memory-Map) or if I have difficulty getting a fix.

That patch - neat find, many thanks. Just downloaded. Just need to get my
smartphone backed up carefully before I install it.

Are there any other known causes of such bias or displacement? It's so
consistent.

FWIW, back in the UK I have on several occasions had weird results from the
same device, as described in this post of 6th April, "Erratic GPS signals".

--
Terry, East Grinstead, UK

QuickGPS is just an alternative way of getting satellite ephemeris and
clock data. It's probably done more to conserve handset battery
energy than for speeding up first fix, though it does that too.

The HTC Touch Pro 2 User Manual indicates that the downloaded data is
valid for nearly 7 days. That requires a definition of the word
"valid".

Currently the GPS Control Segment predicts and uploads to the
satellite, nominally once a day, its ephemeris and clock data for the
next 24 hours, arranged in 2 hour blocks. They also upload longer
term predictions, but that's not too relevant to this thread.

Nominally every 2 hours there is a "page change", when the satellite
starts broadcasting the data that is a best fit for the next 2 hour
time block. If you calculate a position using the "before" and
"after" data, you might see a difference on the order of a meter or
two.

As far as I know, there is no publicly available source for the
Control Segment's predicted data.

The "as broadcast" data can be retrieved within minutes of its first
transmission via various IGS service centers, and probably other
sources too.

So the QuickGPS service provider has some other source of the
predicted ephemeris and clock data in order to get a week's worth of
future data. Not too hard to do it themselves. There have been a
number of patents filed and papers published in the last few years on
methods for creating "extended ephemeris" (Jim Garrison at Purdue,
Lionel Garin from SiRF, etc.). Most or all of these clearly state
that they are predicting only the satellite orbits, not the satellite
clock errors.

The bottom line is that you need to be very skeptical about using old
ephemeris data, just because the display says it's "valid" for another
6 days or whatever.

The ephemeris data is designed to "gracefully degrade" as its age
increases, but "graceful" is a subjective term. It's generally not
even a good idea to use ephemeris data that's more than 2 hours old,
let alone days. But that depends on one's accuracy needs.

At a guess, perhaps on Monday you invoked QuickGPS prior to starting
out, and didn't do it again on Tuesday. If so, the actual ephemeris
and clock data may have changed sufficiently to cause your observed
position errors. It's very unlikely that you did the reverse (update
on Tuesday, no update on Monday).

From skimming the HTC User's Manual, it looks like QuickGPS is the
only means of acquiring ephemeris. It doesn't say anything about the
handset collecting the data from the satellite broadcast signal.

So if feasible and economical (no idea what charges you incur using
QuickGPS), whenever you want the best available accuracy, force a
QuickGPS update prior to setting out.

If there are additional charges for using the service, that might
motivate you to relax your accuracy threshold.

http://htcwiki.wetpaint.com/page/Quick+GPS

Many thanks Ed, greatly appreciate that detailed explanation.

I've now enabled an option in the QuickGPS app that I hadn't noticed before:
'Auto download when data expires'.

Cost should be trivial I think. But what still puzzles me a bit is that I
don't recall serious problems getting an initial GPS signal with my ancient,
less sensitive Garmin GPS 12.

--
Terry, East Grinstead, UK