[time-nuts] Positional accuracy of the M12+T

Magnus Danielson cfmd at bredband.net
Thu Jan 4 04:08:59 EST 2007


From: Dr Bruce Griffiths <bruce.griffiths at xtra.co.nz>
Subject: Re: [time-nuts] Positional accuracy of the M12+T
Date: Thu, 04 Jan 2007 11:03:01 +1300
Message-ID: <459C2815.3070609 at xtra.co.nz>

Björn and Bruce,

> bg at lysator.liu.se wrote:
> > On Wed, January 3, 2007 13:40, Stephan Sandenbergh said:
> >
> >   
> >> The datasheets mention <25m SEP (haven't got a clue what 'SEP' stands for)
> >> positional accuracy. However, I am sure this is relative to the actual
> >> datum. Does anyone know where I could find information on relative
> >> positional accuracy within multi channel common view configurations?
> >>     
> >
> > GPS have strange error metrics, Spherical Error Probability, iirc is 50%
> > of the 3d positions within a sphere  with a radius (diameter?) of 25m.
> > Maybe the metrics come from the military "background" of GPS. SEP says
> > nothing about gaussian error distribution and 50% is even less than 1
> > sigma, whatever that says.
> >
> > Hmmm... with 25m position accuracy (and 100ns is about 30m), how do they
> > really get time down to a few ns. Clever engineering!  :-)
> >
> >
> > --
> >
> >    Björn
> >
> >
> >
> >
> > _______________________________________________
> > time-nuts mailing list
> > time-nuts at febo.com
> > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> >
> >   
> 
> Björn
> 
> 
> The metrics have nothing to do with the military origins of GPS. Similar 
> metrics are used in 3 dimensional surveying. They are a consequence of 
> the 3 (neglecting time) dimensional geometry involved in determining 
> position. A spherical error volume is a crude approximation, actually it 
> is an ellipsoidal with  as the height error is usually significantly 
> larger than the other positional errors which also may have different 
> rms errors. The errors are assumed to have a gaussian distribution with 
> different standard deviations for each coordinate axis. There may also 
> be some correlation between the errors for each axis.
> 
> The concept of spherical error probability seems to have been introduced 
> for those who insist on a single error measure.

For many purposes the heigth value is a mear detail and has no real interest
and then only the horizontal deviation would be of any interest (and here it
should be more accurate). Also, one-sigma measures are good for reference, but
two-sigma or three-sigma values is usually of more interest for those people.
Also, that is somewhat of a worst case number. You are usually better, but as a
good engineer, what value do you promise? :-)

Cheers,
Magnus



More information about the time-nuts mailing list