TangoGPS and GPSD

Marcus Bauer marcus.bauer at gmail.com
Mon Jul 7 13:30:38 CEST 2008

On Sun, 2008-07-06 at 19:52 +0200, Joerg Reisenweber wrote:

> > My Freerunner needs ~5 minutes before it has the time (tangoGPS displays
> > the time of the GPS even if there is no fix) and ~10-15 minutes overall
> > before it gets a fix. If I move, even just walking or have the Neo not
> > in an upright position, I don't get a fix at all. Clear skyview of 80%
> > sky is necessary for my FR too.
> Just to add to and confirm this: acquisition (mere download) of efem/alm takes 
> 12min,

Just to add even more to this: the ephemeris data is repeated every 30
seconds, each satellite sends only its own. The almanac data is repeated
every 12.5 minutes and each satellite sends the data of all satellites.

The important thing is: every modern GPS chip can get a fix without
having the almanac! A "cold start" is a start without initial position,
time, almanac and ephemeris. And all manufacturers claim start up times
~45 seconds. And so does u-blox for their Antaris chips.

In practical life and typical gadets this means ~2min cold start time,
when under good conditions and not moving and ~4min when under good
conditions and walking.

>  if not interrupted by walking around etc, and needs a 30dB higher 
> signal than normal GPS operation.

The datasheet says the needed signal strength is -145dBm for cold start
and -160dBm for tracking. The signal strength of GPS under clear sky is
~126dBm thus the chip needs just 1/100th the signal strength of clear
sky for a cold start.

> Placing the FR flat on a table will probably not see a fix at all, due to 
> orientation of internal GPS-antenna being front long axis.

At least the GTA01 gets a fix inside of my appartment while placed flat
on a table with no sky view at all.

> TTFF can be significantly speed up by supplying time/efem/alm data on startup 
> of GPS-receiver. This can be done by AGPS, or by storing the data locally on 
> device and restoring it to receiver on next power up, given we have correct 
> RTC and didn't move too far since last usage of GPS.

The point of AGPS is to get a fix in as low as 1 second under good
conditions and get a fix in some ten seconds in "urban canyons".

The TTFF time of the Freerunner should be ~2min cold and ~15sec with
AGPS. So far it seems nobody comes anywhere near this.

IMHO it is very irritating that there are no clear answers from the
Openmoko engineers - it seems nobody has ever really tested it and
nobody cares to test it. Instead of theoretical answers I would expect
to get some real world TTFF values.

Even more irritating is that Openmoko engineers believe that 12 minutes
are a 'normal' cold start time.

To add some of my experience here: once the Freerunner gets a fix and
has been running with a fix for ~30 minutes, it still keeps the fix when
going into a brick building and it keeps it at least as long as a decent
SirfIII based GPS. This means it is most likely no hardware problem.

But there are a couple of strong indications that it is a firmware
problem and cannot be solved via the u-blox binary protocol. My guess is
50/50 here.

It would be in Openmoko's own interest to quickly get a clue on this,
otherwise they can expect the first thousand phones to be shipped back
to them. Most likely not so funny...


More information about the openmoko-devel mailing list