Questions For SPOT Gen 3 Owners

[DougPaul]

I have seen sudden jumps in a gps track like you have described (though not 6/10ths of a mile). In my experience, this is because the unit has just switched to a different constellation (in other words - group) of satellites to calculate a position. Think of computing a GPS position as building a house of cards. When a new constellation is chosen, the house falls and the cards have to be stacked up again until they make a stable structure. I hope that is an easier way of saying that a Kalman filter (or any other kind of numerical recipe) has to forget everything it has "learned" and must re-start from scratch.

The contribution of each satellite can be weighted so that satellites can enter and leave the active constellation gradually thus avoiding any steps in position from this cause. And no, you don't need to forget everything whenever a satellite enters or leaves the active constellation.

The other source of the problem comes from marketing departments. Not to open an old debate, but some GPS units claim a much higher sensitivity than other GPS units even though they are very much equal. The "superior" GPS units produce garbage-like noisy positions that drift all over the place when the signal is weak, but as the "inferior" GPS units are designed to report loss of lock instead of garbage-like noisy position data with the same weak signal.

This was a problem with some of the older Magellan units. I don't know if it is still a problem. (I simply haven't read any recent reports on this issue either way about any brand or model.) In any case, when to stop extrapolating and declare a loss of lock is a judgement call.

Doug

 

[DougPaul]

I imagine that there is a somewhat limited value in previous location since Spot cannot tell if, for example, we switched from hiking to driving a car. Nevertheless, last state should be a pretty good starting point for figuring out next location although I am not sure how easy it is to feed into whatever noise-reduction method Spot uses. While Spot has a motion sensor of some sort, I am almost certain it is not used in gps feedback loop as having a decent speed estimate would at least allow throwing away bad readings at step 4.

The location 10 minutes ago would have little effect on the estimate of the current location. (It might only make determination of the new location a little faster to obtain because the GPS may not have to download a new ephemeris. The ephemeris takes a minimum of ~30 seconds to download, is valid for ~4 hrs, and is updated every ~2hrs.)

Consumer GPSes generally receive continuously and produce a new location estimate once per second. Some (many?) Garmin GPSes have a battery saver mode in which they turn the GPS receiver off for ~every other second thus producing a position estimate ~every other second. In one experiment, the RMS error went up by a factor of almost 3 due this mode. http://www.vftt.org/forums/showthread.php?38355-60CSx-battery-lifetimes Tracking the satellite signals (in both time and doppler shift) requires time accuracies on the order of a few nanoseconds--each time the receiver is turned off, the tracking is lost and must be reacquired.

There is no way to tell how many satellites Spot locks on to but my Garmin eTrex 30 almost always reports signal from 7 or more satellites. I guess it does not have a limitation by which it would have to shut itself down every 10 minutes, so it can do a better job of maintaining and re-calibrating its state over time. I often use two of my GPS units (I also have an older Garmin Vista HCX) for the purpose of recording tracks and marking trails on OpenStreetMap. Over time I have compared probably hundreds of miles of track from my two GPS devices, and whenever I walk the same trail back and forth eTrex 30 seems to be a bit more consistent with itself (on average I would say no more than 30 feet of spread between tracks along same path) while the older Vista shows more of a variance (maybe occasionally 70 feet spread) but definitely nothing as big as 6/10ths of a mile.

See http://www.vftt.org/forums/showthread.php?14406-Bakeoff-plus-Tecumseh-hike for some objective (numerical) comparisons of the cross-track repeatability of various GPSes.

IMO, SPOT either uses an obsolete GPS chipset or uses its GPS chipset poorly. Whatever the cause, the bottom line is the accuracy of the SPOT GPS positions is significantly worse than that of a number of modern hiking GPSes.

Doug

 

[Remix]

The contribution of each satellite can be weighted so that satellites can enter and leave the active constellation gradually thus avoiding any steps in position from this cause. And no, you don't need to forget everything whenever a satellite enters or leaves the active constellation.


This was a problem with some of the older Magellan units. I don't know if it is still a problem. (I simply haven't read any recent reports on this issue either way about any brand or model.) In any case, when to stop extrapolating and declare a loss of lock is a judgement call.

Doug

1. With regard to satellites entering or leaving the constellation, please cite a source or some kind of Software Design Document from Garmin. Satellites lower on the horizon cannot produce the same accuracy as satellites at higher azimuths...therefore the position will suddenly have more error. The OP reported ~0.5mi jump. You seriously dont regard this as a step change in position?

2. Actually Doug its a bigger problem with Garmin units.. but Im not going to go around and around and around with you.

 

[DougPaul]

1. With regard to satellites entering or leaving the constellation, please cite a source or some kind of Software Design Document from Garmin. Satellites lower on the horizon cannot produce the same accuracy as satellites at higher azimuths...therefore the position will suddenly have more error. The OP reported ~0.5mi jump. You seriously dont regard this as a step change in position?

AFAIK, Garmin does not publish their internal design documents.

The general method for determining the location is to find the location and time that minimizes the sum of the weighted squares of the differences between the physical distances to the satellites and the pseudoranges* determined from the signals. (There is no closed form solution, but there are a number of iterative methods for solution.) The weights are chosen to reflect the expected error in the pseudorange to the corresponding satellite. A simpler, but less accurate method would be to declare the weights to be 0 or 1 (ie use each satellite equally or totally ignore it).
* The pseuodrange is determined from the time the signal took to travel from the satellite and the GPS.

Sudden loss or acquisition of a satellite is only one possible cause of a step in the estimated position, but it is not the only possible cause. A change in multipath, for instance.... (Using the weighted solution can reduce the impact of some causes of degradation.) Note that modern GPS will generally use more than the minimum of 4 satellites if possible--the additional satellites generally increase the accuracy and improve the ability of the device to deal with a an inaccurate or erroneous pseudorange.

Satellites that are low on the horizon tend to give less accurate pseudoranges than those that are higher up, but if all visible satellites are in a tight group high up you will get large errors. In general, you need a constellation that covers a large portion of the sky for the best accuracy. (See Dilution of Precision (DOP).)

Some references covering much of the above:
http://en.wikipedia.org/wiki/Global_Positioning_System
http://www.edu-observatory.org/gps/gps_accuracy.html
http://en.wikipedia.org/wiki/Receiver_autonomous_integrity_monitoring
http://www.nbmg.unr.edu/staff/pdfs/Blewitt Basics of gps.pdf

FWIW, I am not observing (with a Garmin 60CSx) the sizeable jumps reported by others. In addition, my recorded tracks generally coincide very well with the published FS tracks for the WMNF.

EDIT: I thought about the maximum size of a jump caused by adding or deleting a satellite from the active constellation (assuming 4 or more satellites before and after)--it should be 15 meters or less. A .5 mile jump indicates that something else is going on.

Doug

 

[Remix]

AFAIK, Garmin does not publish their internal design documents.

Sudden loss or acquisition of a satellite is only one possible cause of a step in the estimated

Thats what I said, and I did not address whether there could be other causes.

Satellites that are low on the horizon tend to give less accurate pseudoranges than those that are higher up


Doug

Thats what I said.

In any event, I think there is excessive speculation going on.

The standard way of treating this is to find objective reports regarding performance. Forest Service has published such reports. There are a few individuals who have logged Garmin output versus Magellan Meridian Platinum output while both units are idle under the same signal conditions.

I would rather make decisions and conclusions based on data instead of rampant speculation and hearsay from other discussion groups on the internet.

 

[iAmKrzys]

This youtube video is quite some reminder that even if you hit SOS button on your Spot, or any other PLB for that matter, the help may not arrive until many, many hours later: https://www.youtube.com/watch?v=n-S-9qHYOCs