GPS Inaccuracies and NMEA 1083 protocol

[Joe G \(Home\)]

Hi All,

I am looking for a good indepth forum on GPS and the NMEA 0183 protocol.

What the protocol terms mean "indepth".

Any pointer to a good forum would be appreciated.

Joe

 

[Rich Webb]

Hi All,

I am looking for a good indepth forum on GPS and the NMEA 0183 protocol.

What the protocol terms mean "indepth".

There is no "depth" to the NMEA 0183 standard; it's really very
straightforward.

As regards the combination "GPS and the NMEA 0183 protocol," a given GPS
could emit one or more sentences, periodically or asynchronously, as
determined by the manufacturer.

Did you have a specific question?

 

[Vladimir Vassilevsky]

There is no "depth" to the NMEA 0183 standard; it's really very
straightforward.

I've done NMEA parsers; although the protocol looks straightforward,
there is still a lot of subtleties and a vendor/device dependent things.
It is not very simple to do a universal parser which decodes correctly
all possible strings in all possible variants.

As regards the combination "GPS and the NMEA 0183 protocol," a given GPS
could emit one or more sentences, periodically or asynchronously, as
determined by the manufacturer.

Did you have a specific question?

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[[email protected]]

Hi All,

I am looking for a good indepth forum on GPS and the NMEA 0183 protocol.

What the protocol terms mean "indepth".

Any  pointer to a good forum  would be appreciated.

Joe

You could have a look at the sources of gpsd (Linux).
http://gpsd.berlios.de/
It is not a tutorial, but that page has links to lots of related
stuff.

 

[Rich Webb]

I've done NMEA parsers; although the protocol looks straightforward,
there is still a lot of subtleties and a vendor/device dependent things.
It is not very simple to do a universal parser which decodes correctly
all possible strings in all possible variants.

Particularly since vendors sometimes get things wrong, too. E.g., a
device (unspecified, to protect the guilty) that accepts NMEA data until
*any* "variable number" field (x.x but also hhmmss.ss and lat/lon) fails
to include the radix point with at least one fractional digit. Once an
integer is seen in any one of those fields, then it stops processing
NMEA data.

Arguably permissible if their IDS states a minimum required precision
for those fields (it doesn't), although just failing to operate past
that sentence is not an optimal way to alert the operator. It was a
bastard to figure out, too, since the GPS I was using to drive it filled
the UTC field with integer time-stamps, so it stopped processing NMEA
data as soon as the first NMEA sentence arrived. Grrrrr.

 

[Joe G \(Home\)]

I've done NMEA parsers; although the protocol looks straightforward, there
is still a lot of subtleties and a vendor/device dependent things.
It is not very simple to do a universal parser which decodes correctly all
possible strings in all possible variants.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Yes, I have a specific questions relating to a NMEA 0183 $GPGGA Sentence
on a SiRF StarIII GPS Chipset.

(see GGA extract below)

Questions
======
a) What does [ (6) GPS Quality Indicator = 1] really indicate? More than 4
channel locked and Accuracy is within spec?

b) Is [ 8) Horizontal Dilution of precision ] really indicate metres or a
relative error of measurement ?

c) How do I determine when the NMEA 0183 $GPGGA Sentence will give me
accurate position?


Thanks in advance Joe



Extract from http://gpsd.berlios.de/NMEA.txt
GA - Global Positioning System Fix Data
Time, Position and fix related data for a GPS receiver.

1 2 3 4 5 6 7 8 9 10 | 12 13 14 15
| | | | | | | | | | | | | | |
$--GGA,hhmmss.ss,llll.ll,a,yyyyy.yy,a,x,xx,x.x,x.x,M,x.x,M,x.x,xxxx*hh<CR><LF> Field Number: 1) Universal Time Coordinated (UTC) 2) Latitude 3) N or S (North or South) 4) Longitude 5) E or W (East or West) 6) GPS Quality Indicator, 0 - fix not available, 1 - GPS fix, 2 - Differential GPS fix (values above 2 are 2.3 features) 3 = PPS fix 4 = Real Time Kinematic 5 = Float RTK 6 = estimated (dead reckoning) 7 = Manual input mode 8 = Simulation mode 7) Number of satellites in view, 00 - 12 8) Horizontal Dilution of precision (meters) 9) Antenna Altitude above/below mean-sea-level (geoid) (in meters) 10) Units of antenna altitude, meters 11) Geoidal separation, the difference between the WGS-84 earth ellipsoid and mean-sea-level (geoid), "-" means mean-sea-level below ellipsoid 12) Units of geoidal separation, meters 13) Age of differential GPS data, time in seconds since last SC104 type 1 or 9 update, null field when DGPS is not used 14) Differential reference station ID, 0000-1023 15) Checksum

 

[Rich Webb]

Yes, I have a specific questions relating to a NMEA 0183 $GPGGA Sentence
on a SiRF StarIII GPS Chipset.

(see GGA extract below)

Questions
======
a) What does [ (6) GPS Quality Indicator = 1] really indicate? More than 4
channel locked and Accuracy is within spec?

Probably at least four tracked in SPS mode, although a 2D fix is
possible with three. The precise definition, however, will be vendor
specific and if you need to know more than fix/no-fix you'll have to ask
their tech support.

b) Is [ 8) Horizontal Dilution of precision ] really indicate metres or a
relative error of measurement ?

Relative, based on the geometry of the satellites in use.

c) How do I determine when the NMEA 0183 $GPGGA Sentence will give me
accurate position?

The quality indicator, the number of satellites in use, and an overall
reasonableness check. Actual error estimates are available in the GST
sentence but that's not commonly provided.

 

[Vladimir Vassilevsky]

Questions
======
a) What does [ (6) GPS Quality Indicator = 1] really indicate? More than 4
channel locked and Accuracy is within spec?

It indicates whatever the vendor decided to indicate. You can only think
about it in the terms of "good signal" or "bad signal".

b) Is [ 8) Horizontal Dilution of precision ] really indicate metres or a
relative error of measurement ?

This characterizes the geometric strength of the current constellation
of the satellites. There is no way to make any assumption about the
actual accuracy from "HDOP" and "VDOP" parameters; it is all device
dependent. Again, you can only think in the terms of "better" or "worse".

c) How do I determine when the NMEA 0183 $GPGGA Sentence will give me
accurate position?

If the GPS has the position fix, you will receive it. If it doesn't,
then the GPS will give you the empty fields for latitude and longitude.
The NMEA philosophy is the GPS receiver decides if it has the position
fixed. It is not the business of your application to make any
assumptions or conclusions.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[Nico Coesel]

Questions
======
a) What does [ (6) GPS Quality Indicator = 1] really indicate? More than 4
channel locked and Accuracy is within spec?

It indicates whatever the vendor decided to indicate. You can only think
about it in the terms of "good signal" or "bad signal".

b) Is [ 8) Horizontal Dilution of precision ] really indicate metres or a
relative error of measurement ?

This characterizes the geometric strength of the current constellation
of the satellites. There is no way to make any assumption about the
actual accuracy from "HDOP" and "VDOP" parameters; it is all device
dependent. Again, you can only think in the terms of "better" or "worse".

c) How do I determine when the NMEA 0183 $GPGGA Sentence will give me
accurate position?

If the GPS has the position fix, you will receive it. If it doesn't,
then the GPS will give you the empty fields for latitude and longitude.
The NMEA philosophy is the GPS receiver decides if it has the position
fixed. It is not the business of your application to make any
assumptions or conclusions.

I disagree. In areas with a lot of buildings the error may be well
over 120 meters even thouh the GPS thinks its locked! At my previous
employer I used a GPS to track the position of a verhicle for use in
research on how people drive a car in certain circumstances. The data
(not only GPS) needed to be analysed between various waypoints. The
routines to determine if a piece of data needed to be included in an
analyses wasn't simple.

 

[Rich Webb]

I disagree. In areas with a lot of buildings the error may be well
over 120 meters even thouh the GPS thinks its locked!

And its thinking is correct. That is, the pseudorange measurements that
it is making to the tracked satellites are (presumably) correct and the
position estimate would be correct *if* the signal paths to the
satellites were straight lines.

In an urban canyon, however, the direct path to a satellite may be
obscured, leading the receiver to track a reflected path with better
signal strength. The distance from the receiver to the satellite is
larger, so the sphere of position for (at least) that satellite has a
larger radius than it "should."

The tracker can only assume that it's operating with all the satellites
in direct line of sight. So, when it crunches the numbers and finds the
centroid at the intersection of all of the spheres of position, the
estimated position will be offset by some tens or, possibly, hundreds of
meters from the actual geographic position.

The same receiver at the same location but at a different time will have
different results because the geometry of the constellation will have
changed.

 

[Nico Coesel]

And its thinking is correct. That is, the pseudorange measurements that
it is making to the tracked satellites are (presumably) correct and the
position estimate would be correct *if* the signal paths to the
satellites were straight lines.

In an urban canyon, however, the direct path to a satellite may be
obscured, leading the receiver to track a reflected path with better
signal strength. The distance from the receiver to the satellite is
larger, so the sphere of position for (at least) that satellite has a
larger radius than it "should."

The tracker can only assume that it's operating with all the satellites
in direct line of sight. So, when it crunches the numbers and finds the
centroid at the intersection of all of the spheres of position, the
estimated position will be offset by some tens or, possibly, hundreds of
meters from the actual geographic position.

The same receiver at the same location but at a different time will have
different results because the geometry of the constellation will have
changed.

Exactly. I plotted the data of several drivers taking the exact same
route on top of eachother on a map and in some areas you get an
enormous blur. There is no way to tell what the actual error is.

 

[krw]

And its thinking is correct. That is, the pseudorange measurements that
it is making to the tracked satellites are (presumably) correct and the
position estimate would be correct *if* the signal paths to the
satellites were straight lines.

In an urban canyon, however, the direct path to a satellite may be
obscured, leading the receiver to track a reflected path with better
signal strength. The distance from the receiver to the satellite is
larger, so the sphere of position for (at least) that satellite has a
larger radius than it "should."

The tracker can only assume that it's operating with all the satellites
in direct line of sight. So, when it crunches the numbers and finds the
centroid at the intersection of all of the spheres of position, the
estimated position will be offset by some tens or, possibly, hundreds of
meters from the actual geographic position.

The same receiver at the same location but at a different time will have
different results because the geometry of the constellation will have
changed.

Sure, all of that is apparent. However, I would have thought that
given enough satellites there shouldn't be a solution if one or more
had additional path length inserted. Doesn't the receiver flag that?
Or does it just throw out a satellite with data that doesn't "fit".
Average? It seems that it _should_ fail to lock, though that could be
a marketing disadvantage.

 

[Rich Webb]

Sure, all of that is apparent. However, I would have thought that
given enough satellites there shouldn't be a solution if one or more
had additional path length inserted. Doesn't the receiver flag that?
Or does it just throw out a satellite with data that doesn't "fit".
Average? It seems that it _should_ fail to lock, though that could be
a marketing disadvantage.

Absent other constraints (e.g., odometry, inertial reference, ...) it's
looking, more or less, at a weighted mean position where the weights are
taken from SNR, correlation coefficients, etc. Yes, if only one
satellite of, say, six was affected then it would have correspondingly
less of an effect on the solution. In amongst blocks of tall buildings,
though, it's a bit more challenging since most of the sats could be out
of a direct view.

The processor itself probably knows very well that the error estimate
has grown but the user equipment may not see that, depending on how it
is interfaced to the GPS.

 

[Joe G \(Home\)]

Questions
======
a) What does [ (6) GPS Quality Indicator = 1] really indicate? More
than 4
channel locked and Accuracy is within spec?

It indicates whatever the vendor decided to indicate. You can only think
about it in the terms of "good signal" or "bad signal".

b) Is [ 8) Horizontal Dilution of precision ] really indicate metres or
a
relative error of measurement ?

This characterizes the geometric strength of the current constellation
of the satellites. There is no way to make any assumption about the
actual accuracy from "HDOP" and "VDOP" parameters; it is all device
dependent. Again, you can only think in the terms of "better" or "worse".

c) How do I determine when the NMEA 0183 $GPGGA Sentence will give me
accurate position?

If the GPS has the position fix, you will receive it. If it doesn't,
then the GPS will give you the empty fields for latitude and longitude.
The NMEA philosophy is the GPS receiver decides if it has the position
fixed. It is not the business of your application to make any
assumptions or conclusions.

I disagree. In areas with a lot of buildings the error may be well
over 120 meters even thouh the GPS thinks its locked! At my previous
employer I used a GPS to track the position of a verhicle for use in
research on how people drive a car in certain circumstances. The data
(not only GPS) needed to be analysed between various waypoints. The
routines to determine if a piece of data needed to be included in an
analyses wasn't simple.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
"If it doesn't fit, use a bigger hammer!"
--------------------------------------------------------------

Dear Nico,

Thank you for your email.... can you give some hint how you determined the
position inaccuracies?

Did you use average measurements? Use HDOP as an error indicator?

Any clues would be appreciated.

Thanks in Advance.
Joe

 

[Nico Coesel]

Joe G (Home) wrote:

Questions
======
a) What does [ (6) GPS Quality Indicator = 1] really indicate? More
than 4
channel locked and Accuracy is within spec?

It indicates whatever the vendor decided to indicate. You can only think
about it in the terms of "good signal" or "bad signal".


If the GPS has the position fix, you will receive it. If it doesn't,
then the GPS will give you the empty fields for latitude and longitude.
The NMEA philosophy is the GPS receiver decides if it has the position
fixed. It is not the business of your application to make any
assumptions or conclusions.

I disagree. In areas with a lot of buildings the error may be well
over 120 meters even thouh the GPS thinks its locked! At my previous
employer I used a GPS to track the position of a verhicle for use in
research on how people drive a car in certain circumstances. The data
(not only GPS) needed to be analysed between various waypoints. The
routines to determine if a piece of data needed to be included in an
analyses wasn't simple.

----

Dear Nico,

Thank you for your email.... can you give some hint how you determined the
position inaccuracies?

Did you use average measurements? Use HDOP as an error indicator?

None of the above. I plotted the recorded GPS data from several trips
onto a map afterwards. Like others pointed out: GPS signals can
reflect from buildings. The receiver has no way to tell the
difference. You should take into account that GPS was designed to be
used on battle fields which usually are not areas with a lot of
buildings.