5500 foot 855 startup altitude error

Wow, at that altitude I be too busy trying to catch my breath.
Sorry, I couldn't resist.
Does it correct itself after driving a short distance?
It is possibly not receiving all the satellites before you start moving/driving.
My 765T sometimes takes up to a mile after it is first turned on to get all the info that it wants/needs.

I'd agree serious multipath, and you should know better than to expect a first fix to be close, particularly when you suspect wild multipath issues.

And once you get that first fix, think about what's happening -- that's your starting point for a Kalman filter that's tuned for automotive applications. And for automotive applications, that Kalman filter is going to expect maximum position changes (velocities) in the X and Y directions of tens of meters per second, right? And since you're in a car, it's also going to expect the Z (altitude) velocity to be a lot lower, right? Cars seldom experience sustained Z changes of 10 meters/second -- we hope!

As your GPS finds more birds, it quickly moves from a minimum in-motion solution, 4 birds, to an overdetermined solution, more than 4 birds, and if the code is well designed, that wobbly multipath signal will get all but ignored.

So as the Kalman filter works that bad first fix out of its system, X and Y converge quickly -- at Porsche GT3 speeds! But Z converges slower, because even that GT3 works better when all four wheels are on the ground, and the Kalman filter knows that!

Overall, sounds like a great view of what's going on behind the curtain in your GPS!

--bob in sunny silicon valley

Yes, the speed limit in my neighborhood is only 20 mph and I observe it. I was about 3/10 of a mile down the road by the time the errors became negligible.

With the normal "always on" behavior of the 855, it has usually had a good fix on first startup of the car.

Now that I am trying harder to stop it from depleting my car battery, watching the true cold start behavior is a new thing for me.

Although I probably have actually seen multipath in small amounts fairly frequently, in my nearly 2 decade GPS usage history I've only actually noticed it confidently once before. I carried a GPS V on a trip in the lesser Antilles. The Sea Cloud II was large for a sailing ship (about 4000 tons) but tiny compared to one of the monster (roughly 100,000 grt) cruise ships we pulled alongside in the final docking at Barbados. When I reviewed the track log data later, I saw a nice straight line of fixes about 200 yards inland. I think my GPS had been in the porthole of my cabin at the time, which with the steel hull meant it was always fairly challenged to get four satellites for a fix. While we were traveling along toward our final docking location it probably used as one of its four satellites a signal that was reflected off the cruise ship side, which of course would have it miss locating us by the amount I saw.

It was when I overlaid the track log with Google Earth overhead imagery that I chuckled and saw what had happened. I may be able to find the track log data and post an image at scale suitable to see the effect. I could tell from the match of subsequent track log data from a bus trip around the island that the registration error of the Google Earth overhead imagery to my log was small compared to this episode, though it was perhaps 50 feet--larger than I am used to seeing in the US. For now I'll post a link to the island-scale track log, to see how this forum handles links.

http://i6.photobucket.com/albums/y231/PeterAStoll/trips/Less...

It does not appear this forum supports use of the IMG tag to allow images to appear inline.

The link below goes to a page showing the Barbados multipath incident I attribute to a satellite signal being reflected from the side of a cruise ship.

http://s6.photobucket.com/albums/y231/PeterAStoll/trips/Less...

This track log was acquired by a Garmin GPS V. Possibly the abrupt transitions indicate it used less of the Kalman filtering alluded to by k6rtm than does my 855. It also was not using road lock, which changes things.

I've uploaded the track log from my recent Sandia Crest incident to my PC, but there is more work to do to post a useful view of the data.

As viewed in Mapsource, the initial phase of the multipath error recovery is shown in orange. My actual position was at waypoint "home" while the GPS initially placed me thousands of feet east. After that, as it corrected error every more westerly, the snap to road function caused the locations to (falsely) four successive wrong road segments (not all visible at the scale shown) before finally catching up to my actual position on the road out of our neighborhood.

http://i6.photobucket.com/albums/y231/PeterAStoll/problems/G...

I've left my return trip in blue, to show what the actual road route taken would have looked like.

Uploading the paths to Google Earth permits viewing the event first from a simple direct overhead point of view:

http://i6.photobucket.com/albums/y231/PeterAStoll/problems/G...

and then from a slant view which gives a hint of how the mountain looms close to my initial position:

http://i6.photobucket.com/albums/y231/PeterAStoll/problems/G...

As to the initially described first slow then fast correction of altitude error, here is the set of positions preserved in my track log.

http://i6.photobucket.com/albums/y231/PeterAStoll/problems/G...

I don't think the track log preserves every position calculated (and displayed on the satellite page)--but a subset chosen to give a good path representation. In any case, you can see the pattern of initial small corrections, followed by very large and rapid corrections, somewhat sooner in altitude than in horizontal position. No, I did not exceed the neighborhood 20 mph speed limit by the massive amount shown here.

This next is rather speculative, as I, though an electrical engineer, am not an RF guy, nor an antenna guy, nor a GPS guy. It seems likely that by rare chance that morning a satellite to my west bounced a signal from a section of rock in the Sandia face that functioned as a bit of antenna to some degree focussed on my house. Were Sandia Crest simply functioning as a flat plate reflector, the granite face would have scattered the signal something more like a Lambert reflector, and certainly nowhere near a flat plate. The signal reaching my house would have blurred across a broad range of delays, precluding successful false reception at a particular time.