As the use of SatNav systems increases among the general user population the misconceptions that have arisen on how these systems work are spreading. This article covers one of the author's favourites, the 'fact' or 'common knowledge' that GPS accurately pinpoint you anywhere in the world. This article is aimed at increasing awareness of what is really going on with accuracy.
Well, can anyone define pinpoint accuracy? It must vary as specs for different GPS units state different figures. My new Garmin handheld states less than 10m 95% typical with the proviso 'subject to accuracy degradation to 100m 2DRMS under the US DoD Selective Availability (SA) Program when activated'. I have had it down to saying 3m or even 2m in the higher accuracy differential mode (DGPS). However, whether to believe it or not at these better figures? A survey GPS unit states, Single Point L1 1.8m CEP (most will be explained later).
These figures are very very good, but are they pinpoint?
Let us keep analysing. Perhaps if it is good enough that it doesn't matter, it is pinpoint. If your GPS can get you close enough to use the best navigation system in the world, the Mark I Eyeball, then it has done its job. Depending on light/dark and weather this can range from a few meters to 10 miles or more. Is it pinpoint if it gets you to within 10 miles, where you can see your target, on a clear bright day?
I always thought the definition was, er, the point of a pin, less than 1mm. Only the best GPS left stationary for days will get anywhere near this.
Furthermore many talk of accuracy loosely without understanding. Even I have been loose with the term. There is accuracy, precision or repeatability and resolution, often with cross meanings to different people. Probably that is why many are loose with the term.
When someone says pinpoint I am tempted to ask for a figure and when and how often, resulting in the blank look. This brings in 95%, 2DRMS and CEP. Accuracy is defined as a statistical measure, it is pointless not to. The figure, for example, <10m is maybe the expected under best conditions rather than the average. These best conditions are where in the world there is full coverage and good satellite geometry, their position relative to the receiver. Its all about angle of cut. An example of how often is the 95%, the accepted but not exactly 2 sigma. 95% of recorded positions are expected to be within 10m of the true position. If you wanted 99% that would mean something in the order of maybe 15m. Circular Error Probable (CEP) is 50%.
Now for the anywhere in the world. Due to the orbital inclination you can expect less visible satellites above 55° north and south, less to choose from to obtain the required accuracy position solution. This is not the end for the position is also relative to some reference system and can cause issues if you are combining the use with a map.
GPS works on a reference system that is the best global fit. You may have come across WGS84 Datum. The world has a mathematically horrible shape, all lumpy and bumpy so only a best fit is viable. For some areas this is not a good fit so a different datum is used. Many of these precede GPS. The result is that plotting your WGS84 position on a map of a region of different datum can mean hundreds of meters out. For the car SatNav which is of proprietary manufacture this will be of no consequence as the map would, or should, come already referenced. If navigating with a hand held you need to know the datum for the area. Again the conversion sums are horrible but, no fear, most hand helds have most datums in the software, as long as you know this and where to find them you will be OK.
Essentially that pinpoint is more degraded by plotting on the wrong datum, something not always considered by many.
So there you go. If you arrive safely at your desired destination then the system has done its job. But lets not call it what it isn't.
There are other misconceptions but these are left for a separate article, 'GPS and SatNav. Removing Misconceptions' by the same author.
No comments:
Post a Comment