Date:         Thu, 9 May 2013 11:58:08 -0400
Reply-To:     george jannini <georgejoann@GMAIL.COM>
Sender:       Vanagon Mailing List <vanagon@gerry.vanagon.com>
From:         george jannini <georgejoann@GMAIL.COM>
Subject:      Re: GPS speed accuracy
In-Reply-To:  <vanagon%2013050822461777@GERRY.VANAGON.COM>
Content-Type: text/plain; charset=ISO-8859-1

Thanks David.  I dug the W.E. Coyote reference, having always admired his
fascinating trompe l'oeil tunnel renderings.

Geo/ATL


On Wed, May 8, 2013 at 10:45 PM, David Beierl <dbeierl@attglobal.net> wrote:

> At 04:20 PM 5/8/2013, george jannini wrote:
>
>>    My  Garmin Sat Nav receivers display navigational accuracy between 7
>> feet and infinity (lost in the woods) I can put two of them side by each
>> on
>> the dash, the 2004 model says I'm within 7 feet, but a 2008 model with a
>> much more sensitive chip is always at least double that.
>>
>
> If you seriously want a device that will output Doppler data with its
> track data so you can analyze speed very closely after an event, the
> Locosys GT-31 is your baby, and it's not even expensive.  Real-time
> Doppler speed in a GPS is either completely unavailable or costs
> stupid money.  Furuno has a depth sounder that will do it for
> $80,000, but that's not a GPS.
>
>
> However based on readings I've taken in the last couple hours, a GPS
> unit relying on a single pair of position fixes to determine speed
> could have an error of at least 4 mph for a single reading taken at
> 60 mph.  However the great bulk of the time it would be 1 mph or
> less.  I've found one (marine) unit that quotes an RMS (Root of the
> Mean of the Squares) figure of about +/- .35 knots, meaning the
> reading will be within that most of the time (66%?  I forget).  I
> think that +/- half a statute mile per hour the great bulk of the
> time would be fair for automotive GPS.
>
>
> There is always jitter in the signal.  If you want to see it, set
> your receiver to track movement with a recording interval as often as
> possible and leave it in one place for an hour or two.  Many
> automotive-type GPS units may not show this because they are
> programmed to latch onto the nearest road and if they're within a
> couple hundred feet of a road will show you as being on it.  To get
> them to display jitter I suspect taking them far from any roads or
> trails would work.  I've got my Garmin GPSmap 76 collecting some data
> now, with WAAS enabled, using an external antenna.  With claimed
> accuracies mostly from 6.8 to 7.6 feet, I'm finding each successive
> fix is somewhere between 0.5 feet and 1.6 feet different from the
> previous one.  So far, after an hour or so of tracking, the various
> excursions could be covered by a fifty-foot circle.
>
> Well now.  Even as I finished writing that, our friend took a hike to
> the ESE at a maximum of 4.5 feet between fixes with claimed accuracy
> of 10 feet.  It would now take a 100 foot circle to cover his
> perambulations.  He then went roughly due north and is gradually
> heading WSW to slowly get back to his previous haunts; but he seems
> to have taken up residence for the moment about 35 feet west of them
> and is starting a new position cluster there. 1 to 2.2 feet between fixes.
>
> Three different runs of roughly 1,000, 2,000, and 3,000 seconds each
> produced a track that could be covered by a 100-foot circle, with
> most of the readings occurring within +/- 25 feet of the "center of
> mass".  I do not have any way to determine the absolute positional
> accuracy of the fix but the center seems to be within fifty feet of
> the true position.
>
> Next, there is a factor called HDOP, or Horizontal Dilution Of
> Position (also VDOP but it's of less importance to us).  This is a
> figure of merit for loss of accuracy in the system caused by the
> geometrical configuration of the various satellites visible to the
> receiver at a given time.  Therefor at certain times of day in your
> location, the position accuracy of the system will be degraded
> because of the need to work with small angles (same reason the
> altitude measurement is only specified to within +/- 100 feet, the
> vertical angles are much smaller than the horizontal ones).  I've
> been sitting here watching the HDOP go up and down between about one
> and four, and watching the GPS claimed accuracy roughly tracking it
> between 8 and 23 feet.
>
> Third, there is a systematic error introduced by the GPS receiver and
> its software.  If you take five different GPS units of different
> models and manufacturers and place them in the same location for
> several days, logging the positions they report, you will find that
> some of them have quite a bit of variation compared to others.  We
> would say the second sort are more precise.  However if you then
> average the logged positions, you will probably find that no one of
> the five gives the exact same position, and that a unit with large
> variations nonetheless may give a more accurate position over time
> than one with small minute-to-minute variations.
>
> Fourth, at least with automotive GPS, if the device loses signal I
> believe it will attempt to perform dead reckoning for a little while
> in the hope that it will see some satellites before you notice there
> was a problem.  Not sure whether marine units do this as well but I'm
> inclined to think they do.
>
> Fifth, the displayed is always averaged over the last several
> readings, which will reduce the effect of outliers.  On my marine
> unit I can select the averaging period down to a minimum of two
> seconds, or I can let the GPS decide for itself.
>
> All these things affect the accuracy of the speed reading you see on the
> GPS.
>
> Now some more geometry, that of the land.  If you jump off a cliff,
> will the GPS show you accelerating to your date with Wile E. Coyote,
> or will it show zero?  That would depend on how fancy the software
> people wanted to get.  However at the very least the accuracy would
> be degraded because the altitude measurement is inherently
> inaccurate.  I would lay (small amounts of) money on the units not
> compensating for slopes.  This could be tested by an extremely adroit
> skier with a GPS and a radar gun. ;-)
>
> Yours,
> David
>