How is Position Accuracy Calculated?

[myrrh]

How does my eTrex Vista calculate the accuracy figure shown on the satellite page? In other words, how does it know the current accuracy is (for instance) 23 feet?

[+Red90]

The formula that Garmin uses has not been made public, so the answer is we do not know....

 

Traditionally EPE (Estimated Positional Error) is a percentage probability of being with a radius of the actual location. It uses satellitle signal strength and positional geometry to determine what is the possible error band. There will also be some addition for average atmospheric signal disturbance depending on whether WAAS is used of not.

[+trainlove]

That's the million dollar question.

No manufacturer of GPS's ever describes their algorithm for their secret definition of EPE Estimated Position Error, or DOP Dilution of Precision (H or V).

Just know that the smaller the number the better.

And that having a 3D position fix is better than a 2D one.

And that having a WAAS correction is better than not having it.

Edited October 16, 2008 by trainlove

[+StarBrand]

Do not put too much stock into that number. Try to read it as something like:

 

"I estimate the current calculated position to be about accurate to within xx feet - at least 60% of time under similar conditions (sat configuration, signal strength, etc). But I don't really know or I'd give you better coordinates"

[+johnling]

There will also be some addition for average atmospheric signal disturbance depending on whether WAAS is used of not.

 

I just looked this up the other day. WAAS is supposed to give you an accuracy of 3 meters or better no matter what the GPS displays as accruacy.

[+julianh]

There will also be some addition for average atmospheric signal disturbance depending on whether WAAS is used of not.

 

I just looked this up the other day. WAAS is supposed to give you an accuracy of 3 meters or better no matter what the GPS displays as accruacy.

Absolutely NOT!!!

 

WAAS can give you POTENTIAL accuracy of around 3 metres PROVIDED you have good satellite reception conditions (no multi-path, no tree-cover, etc) AND the satellite constellation above you is favourably configured AND you hold the unit such that the antenna is optimally oriented THEN you should be able to get approximately 3 metre precision MOST of the time. If any of the above are not totally attained, then your achievable accuracy will be worse - sometimes MUCH worse, even with WAAS.

 

GPSr manufacturers don't specify what the EPE really means, but the available evidence seems to suggest that it is the unit's internal estimate of the 50% precision or thereabouts, based on the quality of the signals it is currently receiving, and the disposition of satellites above. That is, the unit estimates that you have a 50% chance (say) of being within the EPE distance, and therefore a roughly 50% chance of being further than the EPE distance. Because this is a statistical calculation, there is actually no upper limit to how big the actual error could be - it could easily be 2 or 3 times the displayed EPE, or even worse.

 

I am sure we have ALL seen occasions when the unit tells us we have an EPE of say 4 metres, but we KNOW we are actually 10 metres or more away from GZ. Conversely, MOST of the time, with good reception, you may see an EPE of say 5 metres, but your real-time accuracy may actually be within 1 or 2 metres. Sadly, there is just no way of knowing (unless you are standing at a surveyed benchmark!)

 

To be statistically meaningful, the manufacturers should tell us whether the figure is the 50% estimate (or whatever), and preferabaly they should really use something like the 95% estimate. However, the 95% estimate would look a LOT worse to the average consumer than the 50% estimate, and so to maintain market competitiveness, they like to imply accuracy that is better than is actually achievable. For personal navigation, it doesn't really matter whether it is the 50%, 60%, 90% or 95% figure - just treat the EPE figure as an estimate of whether your current location accuracy is likely to be "good", average" or "poor" - and treat it with a grain of salt!

 

I don't believe there is any absolute specification of attainable precision in consumer GPSrs under typical real-world conditions, with or without WAAS.

Edited October 17, 2008 by julianh

[+Klemmer]

I've always maintained that the marketing department of a GPSr manufacturer has a major input into the formula for EPE. They certainly don't want to see a large number!

[+johnling]

There will also be some addition for average atmospheric signal disturbance depending on whether WAAS is used of not.

 

I just looked this up the other day. WAAS is supposed to give you an accuracy of 3 meters or better no matter what the GPS displays as accruacy.

Absolutely NOT!!!

 

WAAS can give you POTENTIAL accuracy of around 3 metres PROVIDED you have good satellite reception conditions (no multi-path, no tree-cover, etc) AND the satellite constellation above you is favourably configured AND you hold the unit such that the antenna is optimally oriented THEN you should be able to get approximately 3 metre precision MOST of the time. If any of the above are not totally attained, then your achievable accuracy will be worse - sometimes MUCH worse, even with WAAS.

 

GPSr manufacturers don't specify what the EPE really means, but the available evidence seems to suggest that it is the unit's internal estimate of the 50% precision or thereabouts, based on the quality of the signals it is currently receiving, and the disposition of satellites above. That is, the unit estimates that you have a 50% chance (say) of being within the EPE distance, and therefore a roughly 50% chance of being further than the EPE distance. Because this is a statistical calculation, there is actually no upper limit to how big the actual error could be - it could easily be 2 or 3 times the displayed EPE, or even worse.

 

I am sure we have ALL seen occasions when the unit tells us we have an EPE of say 4 metres, but we KNOW we are actually 10 metres or more away from GZ. Conversely, MOST of the time, with good reception, you may see an EPE of say 5 metres, but your real-time accuracy may actually be within 1 or 2 metres. Sadly, there is just no way of knowing (unless you are standing at a surveyed benchmark!)

 

To be statistically meaningful, the manufacturers should tell us whether the figure is the 50% estimate (or whatever), and preferabaly they should really use something like the 95% estimate. However, the 95% estimate would look a LOT worse to the average consumer than the 50% estimate, and so to maintain market competitiveness, they like to imply accuracy that is better than is actually achievable. For personal navigation, it doesn't really matter whether it is the 50%, 60%, 90% or 95% figure - just treat the EPE figure as an estimate of whether your current location accuracy is likely to be "good", average" or "poor" - and treat it with a grain of salt!

 

I don't believe there is any absolute specification of attainable precision in consumer GPSrs under typical real-world conditions, with or without WAAS.

 

WAAS removes the the intentional errors inserted into the GPS signal. The WAAS correction is transmitted from 25 ground station across the US and is only available in the US. It does not amplify the signal or change the sensitivity of the GPSr unit. It's that simple.

[+Klemmer]

johnling:

Thanks to ex-president Clinton, there have not been any intentional errors inserted into the GPS signal since around May 2000, previously known as Selective Availability (SA).

Edited October 17, 2008 by Klemmer & TeddyBearMama

[+StarBrand]

...

WAAS removes the the intentional errors inserted into the GPS signal. The WAAS correction is transmitted from 25 ground station across the US and is only available in the US. It does not amplify the signal or change the sensitivity of the GPSr unit. It's that simple.

uh - no.

 

WAAS - Wide Area Augmentation System. They are separate Satellites that transmit corrections that are well measured by the more than 25 well surveyed ground stations. Mostly accounting for atmospheric errors and slight corrections to the satellite positions. It uses a mathamatical model to estimate the needed corrections for the area you are in. Actually works well outside of the US in some areas and there is a similar system in use by the European sats. While it can increase accuracy down to 3 meters - it isn't designed to be much more precise.

 

There is no longer any built in errors to GPS - survey grade units (around $4000) can easily get accuracy down to 2 or 3 inches.

[+DENelson83]

BTW, most of those well-surveyed ground stations correspond to so-called "Air Route Traffic Control Centers", owing to the fact that WAAS is the FAA's baby.

 

Also, WAAS being only available in the US? Puh-leeze! If you check this page, you'll see WAAS availability in Canada, Mexico, and even Cuba. That's because there aren't just 25 WAAS reference stations, but 38. Five of them are in Mexico, four in Canada, and the rest are in the US.

 

I should know, because I'm in Canada, and I've actually gotten accuracy on the order of about one metre on my GPSr using WAAS.

Edited October 17, 2008 by DENelson83

[+julianh]

WAAS removes the the intentional errors inserted into the GPS signal. The WAAS correction is transmitted from 25 ground station across the US and is only available in the US. It does not amplify the signal or change the sensitivity of the GPSr unit. It's that simple.

johnling,

 

That's not what WAAS does at all. As Klemmer points out, the "intentional errors" you refer to (known as "Selective Availability" or SA) have not been used for about 8 years now. When SA was in use, land-based DGPS was required for users who wanted higher accuracy by eliminating those deliberate errors (and would also help to correct for system errors in the process), but WAAS is all about correcting for the other system errors which are inherent in the way GPS works.

 

You might want to follow this link for a reasonably succinct description of how WAAS works:

 

http://en.wikipedia.org/wiki/Wide_Area_Augmentation_System

 

You will note that it states that:

 

The WAAS specification requires it to provide a position accuracy of 7.6 meters or better (for both lateral and vertical measurements), at least 95% of the time. Actual performance measurements of system at specific locations have shown it typically provides better than 1.0 meters laterally and 1.5 meters vertically throughout most of the contiguous United States and large parts of Canada and Alaska. With these results, WAAS is capable of achieving the required Category I precision approach accuracy of 16 m laterally and 4.0 m vertically.

 

Note that this is talking about very sophisticated aviation systems, NOT consumer handhelds. As far as I know, there is no specification for the requirements for accuracy of WAAS-enabled consumer GPSrs.

[sokolo0]

About EPE - look here: http://forums.Groundspeak.com/GC/index.php...=184158&hl= post from May 14 2008.

WAAS ( EGNOS in Europe) calculates only shift in your position. Value and direction of that shift is calculated and transmitted from ground station to WAAS/EGNOS satellites. When your gps receiver get it in gps signal and "moves" your position according to that calculation, you have "D" on satellite page.

 

Christopher

Edited October 17, 2008 by sokolo0

[+twolpert]

WAAS removes the the intentional errors inserted into the GPS signal. The WAAS correction is transmitted from 25 ground station across the US and is only available in the US. It does not amplify the signal or change the sensitivity of the GPSr unit. It's that simple.

There is no relationship between WAAS and the military's (no longer used) Selective Availability program.

[+trainlove]

...

WAAS removes the the intentional errors inserted into the GPS signal. The WAAS correction is transmitted from 25 ground station across the US and is only available in the US. It does not amplify the signal or change the sensitivity of the GPSr unit. It's that simple.

uh - no.

 

Of course StarBrand is right.

It's just that us possessive capitalist Americans, who invented both GPS and SBAS, tend to call the augmentation system, all of them WAAS instead of SBAS which they technically really are.

There are EGNOS satellites for the European regions, and MSAS for Japan and one yet to be launched for India...

 

Does anyone, perhaps DENelson83, have the actual coordinates for the 'well surveyed' ground stations for the American WAAS stations? I've seen maps of them, I've seen lists that are not complete, some don't list 25 and some don't list 38 but slightly smaller numbers, and not very precise coordinates if I remember right.

 

P.S. If you are not within the 5x5 degree grid covered by a ground station, you sight get some WAAS, oops there I go again I mean SBAS corrections. You might get the Fast and the Long corrections but you will not get the Iono corrections.

Edited October 17, 2008 by trainlove

[yogazoo]

Edited October 17, 2008 by yogazoo

[+Night Stalker]

I have noticed that the EPE on my 60CX is more conservative than on my older 60CS. My older unit in a side by side would consistently show and EPE of 6-9 while my 60CX would show an EPE of 11-15. In most cases the 60CX was as close to be right on as the older 60CS.

[+trainlove]

My older unit in a side by side would consistently show and EPE of 6-9 while my 60CX would show an EPE of 11-15. In most cases the 60CX was as close to be right on as the older 60CS.

If all else is the same reguarding almanac lifetime and geometry and so on, then this is proof that EPE is just a made up number and different versions of firmware just say what they want to say when they convert DOP to EPE. Perhaps turn on NMEA and collect, simultaneously, the position sentences that include DOP from both units and observe if the units actually simul;taneously say different EPE numbers, but actually say the same DOP numbers in your collected data file.

[+apersson850]

There is no ambiguity with DOP. In a perfect world it's 1 (one), but in reality always more. The error estimated from the signal quality should be multiplied with DOP to get the real error.

 

The fact that two different units, the 60 CS and the 60 CSx, comes up with different EPE does not proove it's a made up number. These units have different GPS rceivers, and the significantly higher capability of tracking weak signals in the 60 CSx may make it see problems with the signals of which the 60 CS are unaware, but may still be affected.

[+GPSlug]

WAAS removes the the intentional errors inserted into the GPS signal. The WAAS correction is transmitted from 25 ground station across the US and is only available in the US. It does not amplify the signal or change the sensitivity of the GPSr unit. It's that simple.

There is no relationship between WAAS and the military's (no longer used) Selective Availability program.

Well, to be fair to johnling, WAAS was designed for removing SA as well as the unintentional orbit, clock and iono errors. The amount of correction it can handle is much bigger than it needs now, and actually could get better accuracy if it was redesigned with higher resolution with no SA. But it's also more designed for integrity than it is for accuracy, so that won't be enough of a priority to develop and re-certify new messages.

 

Of course StarBrand is right.

It's just that us possessive capitalist Americans, who invented both GPS and SBAS, tend to call the augmentation system, all of them WAAS instead of SBAS which they technically really are.

IIRC, "WAAS" was coined first and spread around the industry when that was all there was and only later the generalized term "SBAS" was coined to also encompass other systems built on the same standards. Just like "GPS" was generalized to "GNSS" to encompass all the Global Navigation Satellite Systems. I guess it's kind of like using the most popular name brand, like "Kleenex" .

 

Does anyone, perhaps DENelson83, have the actual coordinates for the 'well surveyed' ground stations for the American WAAS stations? I've seen maps of them, I've seen lists that are not complete, some don't list 25 and some don't list 38 but slightly smaller numbers, and not very precise coordinates if I remember right.

 

P.S. If you are not within the 5x5 degree grid covered by a ground station, you sight get some WAAS, oops there I go again I mean SBAS corrections. You might get the Fast and the Long corrections but you will not get the Iono corrections.

You've probably seen this map. The coordinates aren't really relevant. It's called Wide Area for a reason. Local effects around the stations get washed out when adjusted into the system-wide model, so it doesn't really matter how far you are from an individual station.

 

The ionosphere is modeled as grid points on a big shell high up in the atmosphere. To get the correction, the receiver calculates where its line of sight to a satellite will pierce that shell and interpolates between the grid points. If your pierce point is in the grid, you can make an iono correction. It's just that at the edges of the grid, it won't be as good. See here where the GIVE indicates the accuracy of the correction for different parts of the grid. The dots represent where your pierce points are, rather than your location. So near the edge, where you're likely outside the reference station network, you may have some satellites in the good part, some in the not so good part, and some outside. You have to weight the ranges differently and account for a possible bias on the uncorrected satellites, but you can still use the corrections.

[+GPSlug]

There is no ambiguity with DOP. In a perfect world it's 1 (one), but in reality always more. The error estimated from the signal quality should be multiplied with DOP to get the real error.

There's no ambiguity in how DOP is calculated, but there is in how it translates into error. Just scaling the DOP is making an assumption that all the ranges have the same level of error and are weighted the same in the position solution.

 

A more representative error, like the EPE, takes into account that you weight the ranges differently based on the estimated error provided in the ephemeris (or SBAS corrections), the elevation (atmospheric models have more error on low satellites and multipath is more likely), signal strength, etc. Whether it's a 50% number or whatever, the estimated error can be calculated differently for different receivers because of the assumptions made about the external errors and the quality of the tracking chip and antenna.

 

The hardest error to model is multipath. I generally assume that with no reflectors around, the EPE is probably conservative, and I trust it less and less the more possibility there is for multipath or, worse, pure reflected signals where the direct signal is blocked.

 

I think the biggest part of the EPE calculation being "proprietary" is that they don't want consumers or reviewers to be able to test it too closely and give them grief when they weren't able to estimate it well, or to give fuel to their competitors who would make comparisons, fairly or unfairly.

[+DENelson83]

Does anyone, perhaps DENelson83, have the actual coordinates for the 'well surveyed' ground stations for the American WAAS stations? I've seen maps of them, I've seen lists that are not complete, some don't list 25 and some don't list 38 but slightly smaller numbers, and not very precise coordinates if I remember right.

Unfortunately, the coordinates of those stations are classified.

[+coggins]

Unfortunately, the coordinates of those stations are classified.

Are you sure about that?

[+DENelson83]

Unfortunately, the coordinates of those stations are classified.

Are you sure about that?

That's what the FAA told me.

[+coggins]

Unfortunately, the coordinates of those stations are classified.

Are you sure about that?

That's what the FAA told me.

Then someone is confused as they have all 114 locations posted on a federal website.

[+user13371]

WAAS ground station locations are not classified.

 

See page 113, Table 11.1 of this document: WAAS Performance Analysis Report

 

"... shows the WAAS antenna positions as of 4/29/08 (US and MX are IRTF-2000 from OPUS, Canadian sites are IRTF 2005 from CSRS, for WAAS purposes the difference between IRTF-2000 and IRTF-2005 is negligible)."

 

The "114 locations" mentioned by coggins refer to antenna, 3 at each site. There are 38 ground stations.

 

PS: To see the precise locations in Google Earth or Google Maps, look here:

http://bbs.keyhole.com/ubb/showflat.php/Ca.../page/0#1249462

Edited October 19, 2008 by lee_rimar