Position accuracy and conversion between Latitude/Longitude and OSGB36

[Ron Bell]

I am using GPS Essentials on a Samsung Galaxy Note 3 (Android). If I set "Position Datum" to "World Geodetic System 1984" and "Position Format" to "Decimal" and then create a waypoint from my current position (say the summit of Fuar Tholl, a mountain in Scotland where I was on Tuesday), I get Latitude and Longitude values as follows: N57.485378 W-5.379152. If I now change Position Datum to "Ordnance GB 1936, Mean (7 Para)" and Position Format to "OSGB36", the same waypoint (not re-measured) is now shown as: NG 97548 48943. This is, in fact, a very accurate transformation - it is very much better, for example than the old Garmin Etrex H hand held GPS that I used to own. However, it is not exactly correct. If you go to http://www.ordnancesurvey.co.uk/gps/transformation and use the full OSTN02 transformation on that Lat/Long, you get NG 97546.009  48946.691. This is about 4 metres different - a very small amount, I agree, but not exactly the same. I have three questions, please:-
1. What transformation does GPS Essentials use between Latitude/Longitude and OSGB36 and is this standardised across other GPS apps and dedicated GPS devices these days?
2. Do the exact figures you get from Latitude/Longitude (WGS84) depend on the app (GPS Essentials in this case) or is it absolute in the sense that, given the information the device receives from the satellites, do the rules for WGS84 fix it precisely and all GPS apps should give the same values?
3. Is there any reason why Latitude/Longitude figures from a dedicated GPS device might be more accurate than those given by GPS Essentials on my Note 3?
Many thanks.

--

GPS Essentials Version: 4.0.40

Android Version:

Device Model: Samsung Galaxy Note 3

Device ID: (If you uploaded bug reports this helps to associate them with your question or report. If you have security concerns, send to sup...@mictale.com)

[Michael Schollmeyer]

Hi Ron,

> 1. What transformation does GPS Essentials use between Latitude/Longitude and OSGB36 and is this standardised across other GPS apps and dedicated GPS devices these days?

GPS Essentials uses GEOTRANS which is a library originally published by the USGS, I ported it to Java. GEOTRANS calls OSGB "BritishNationalGrid" but I would guess it yields the same results (down to a sub-meter-level) compared to other OSGB<->lng/lat transformations. Another thing that influences the transformation are the datum parameters and the algorithm used to perform the datum shift. Then again depending on these parameters GEOTRANS chooses the algorithm to perform the datum shift (either 3-step-method or Molodensky shift).

This is not carved in stone, though. I am planning to switch to Proj4 because of it's better support for custom coordinate systems. Anyway, IMHO all (valid) implementations of OSGB should be accurate at a cm-level.

May first guess would be that your web site uses a different datum.

> 2. Do the exact figures you get from Latitude/Longitude (WGS84) depend on the app (GPS Essentials in this case) or is it absolute in the sense that, given the information the device receives from the satellites, do the rules for WGS84 fix it precisely and all GPS apps should give the same values?

The GPS receiver derives the location (lng/lat) from the satellite signals and then the OS (Android) reads out the values from the GPS unit and provides the information to all apps. If you have multiple GPS apps reading your current GPS location, then all apps will most probably provide identical values for your longitude and latitude.

I say "most probably" because an app may attempt to increase the accuracy by adding information from other sensors to establish an inertial navigation platform. If you are not moving, an inertial navigation platform cannot add accuracy and all apps should yield identical results. You frequently see reviews that either up- or downvote an app for it's location quality or time-to-first-fix. To put it frankly, this is a misconception about how GPS works on Android.

GPS Essentials is not implementing an inertial navigation platform (yet)

> 3. Is there any reason why Latitude/Longitude figures from a dedicated GPS device might be more accurate than those given by GPS Essentials on my Note 3?

Yes. There are (al least) two other things that influence GPS accuracy: The first is your GPS antenna configuration. E.g. I have a Nexus S and a Nexus 5 here and most of the time the Nexus 5 sees more GPS Satellites. More satellite signals also means increased accuracy. The Nexus 5 also supports GLONASS so on top of the GPS satellites, the device can use these to further increase accuracy. GPS location quality differs dramatically between Android device models, unfortunately no-one seems to make good tests covering this issue.

Cheers, Michael
mictale.com/ms
--

On 27 Jun 2014, at 11:20, Ron Bell <theb...@silverhow.me.uk> wrote:

> I am using GPS Essentials on a Samsung Galaxy Note 3 (Android). If I set "Position Datum" to "World Geodetic System 1984" and "Position Format" to "Decimal" and then create a waypoint from my current position (say the summit of Fuar Tholl, a mountain in Scotland where I was on Tuesday), I get Latitude and Longitude values as follows: N57.485378 W-5.379152. If I now change Position Datum to "Ordnance GB 1936, Mean (7 Para)" and Position Format to "OSGB36", the same waypoint (not re-measured) is now shown as: NG 97548 48943. This is, in fact, a very accurate transformation - it is very much better, for example than the old Garmin Etrex H hand held GPS that I used to own. However, it is not exactly correct. If you go to http://www.ordnancesurvey.co.uk/gps/transformation and use the full OSTN02 transformation on that Lat/Long, you get NG 97546.009 48946.691. This is about 4 metres different - a very small amount, I agree, but not exactly the same. I have three questions, please:-

> Many thanks.
> --
> GPS Essentials Version: 4.0.40
> Android Version:
> Device Model: Samsung Galaxy Note 3
> Device ID: (If you uploaded bug reports this helps to associate them with your question or report. If you have security concerns, send to sup...@mictale.com)
>
>

> --
> You received this message because you are subscribed to the Google Groups "GPS Essentials" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to gpsessential...@googlegroups.com.
> To post to this group, send email to gpsess...@googlegroups.com.
> Visit this group at http://groups.google.com/group/gpsessentials.
> For more options, visit https://groups.google.com/d/optout.

[Ron Bell]

Michael, Thank you very much for your rapid and very interesting reply.

A couple of comments, please:

1. With respect to GEOTRANS and OSTN02 differences, you say "My first guess would be that your web site uses a different datum." The site I quoted is the official UK Ordnance Survey site and it is my opinion that OSTN02 is today the "official" OSGB36 transformation and is not quite the same as GEOTRANS with deviations typically up to about 4m (as in the example I gave). OSTN02 does not compute the transformation; it is a giant look-up table giving offsets for every point over land (undefined for most points at sea). Although I don't understand it fully, the discussion at http://lists.maptools.org/pipermail/proj/2006-January/002031.html more or less says what I just did I think!! This is mainly of academic interest - GEOTRANS is clearly accurate enough for most purposes - and, if I need greater precision, I'll use Lat/Long.

2. I am into providing the Database of British Hills people with accurate10 figure summit references and I lost my Garmin Etrex H recently, which is why I am concerned with the accuracy of Note 3 (I would prefer not to have to buy another dedicated device). So far, I have compared measurements for 3 British mountain summits with figures already in the Hills Database (these are not necessarily exactly accurate, of course). The results are as follows
Hill id, My Latitude, Database Latitude, My Longitude, Database Longitude.
1450, 55.901672, 55 901671, 5.979347(W), 5.979362(W)
1449, 55.902378, 55.902405, 6.004114(W), 6.004122(W)
0984, 57.485378, 57.485385, 5.379152(W), 5.379188(W)
On the whole, these seem to be very close (often to within a metre or 2 - as good as my Garmin used to be) so that I conclude that any difference between antenna or whatever is insignificant - at least in the sort of good open conditions you get on mountain summits. My main worry here is that all my Longitude figures are (slightly) LESS than the database figures - indicating a possible systematic error that way. However, a sample of three is not really big enough from which to reliably draw this conclusion - I clearly need to get myself up some more hills!

Thanks again. GPS Essentials is a great app.