Large offset in series of location fixes

Hello,

I'm hoping someone will be able to help me by explaining the technical
details behind an effect I'm observing with my GPS unit.

It's a Trimble Placer GPS 455, *without* the dead reckoning heading
sensor attached.

The file at http://www.cl.cam.ac.uk/~jjd27/gps-deviation.gif shows the
GPS fixes I got on two identical journeys at different points in time,
travelling from the east side of the image towards the west.

For the rest of the journey (not shown) the blue and red traces are
pretty much on top of each other, as I would expect. When travelling
north up a tree-lined road on the blue journey, few fixes were
obtained, but the fixes that were obtained were always in the correct
place.

However, on the red journey, no fixes were obtained on the tree-lined
road, and the next set of fixes were about 150m south of where they
should have been. And then, after travelling along for a while,
suddenly, as if the GPS unit realised its mistake, the red journey
snapped back to where it should have been all along.

(Along both journeys there were always between 3 and 5 satellites in
view contributing to the readings.)

I'd be grateful if someone could explain what is the cause of this. Is
this a common occurrence? If I had had a dead reckoning unit attached,
I would have understood why the shape of the red trace was correct
despite the incorrect absolute position. And if it was just a case of
getting bad reflections on the red journey, I would have expected this
to have been resolved much earlier rather than it did...

Thanks,

Jonathan

Jonathan Davies wrote:

Do you know if the unit was calculating a '2D' or a '3D' solution
during the period of the discrepancy? In addition to the multipath
possibility you mention, another cause for that kind of offset would be
if the unit only had a '2D' fix and was assuming an incorrect altitude.
So at the beginning of the offset it may have gotten a rather poor
quality 3D fix that gave it a wrong value for the altitude and if
subsequent fixes were all 2D it would continue to use that value in
calculating positions resulting in all the positions being offset from
their correct values. When it finally got a good 3D fix again it would
correct the altitude value and eliminate the offset.

peter wrote:

Thanks for your thoughts. It certainly sounds like a plausible
explanation!

Unfortunately I wasn't logging whether the fixes were 2D or 3D fixes,
but I did record the number of satellites in view and the altitude data.

If I were getting a sequence of 2D fixes, should I expect the altitude
to continue to vary? It does not by any means stay constant along the
erroneous "offset" journey.

And there were several occasions along the erroneous journey where
there were 5 satellites visible, which is presumably enough to give a
3D fix. What is it that causes the unit to decide to only compute a 2D
fix?

Thanks,

Jonathan

Jonathan Davies wrote:


In general I'd expect the reported altitude to remain constant in that
situation. But depending on how the filtering works on that receiver
there could still be some changes in the altitude even if it doesn't
have consistent 3D reception. My Garmins are slow to respond to
altitude changes when the reception is marginal and will frequently lag
behind the true elevation.

But if the recorded altitudes look about the same for the track in both
directions then it's unlikely that this accounts for the horizontal
discrepancy.

Reception of at least 4 satellites is required for a 3D fix. But
depending on the geometry this won't always be sufficient. E.g. if all
the satellites are at the same angle above the horizon then the unit
won't be able to get a good altitude reading.

Jonathan Davies wrote:

The interesting thing about your result is how the error is a simple
lateral shift. I don't think it could be explained with error due to
the number of birds in sight. You would expect the position error to be
on either side of the true course.