I don't know if this is old news or not, but I thought I'd post anway.

I had two N810s that refused to fix (>12 hours) and were sitting on the same window sill as two others that were happily locked on. The two sick ones couldn't see any satellites whilst the other two could see seven or eight with at least four lit up in green.

I found that I could copy the /var/lib/gps/nvd_data from one of the good ones to the two ailing ones, restart the gpsd (via the GPS Enable tick box) and they both recovered.

There is now a script sitting on all four that, should I get stuck again, will copy a 'good' nvd_data to /var/lib/gps. Hopefully this will allow me to get working again if the problem reappears (which it will).

My guess is that, if gpsd shuts down when it is not locked, then garbage can get written to nvd_data. Sometimes it can cope, other times it can't and the effects are unpredictable.

I imagine that, for this workaround to be of use to anyone else, you would need your own 'good' nvd_data saved at a time when you know it is locked on in your part of the world.

HTH

Paxton

P.S. I've tried copying the nvd_data to a fifth N810. My window sill isn't big enough for that one, so it is on a bench well away from the window. It has locked on no problem.

Does removing the nvd_data and restarting the gps have the same/similar effect?

No - I tried that first and it didn't work at all. It looks as though gpsd needs the contents of a good nvd_data. Maybe if you wait for days gpsd will work it out for itself, but Ive not tried that.

Hmmm, I don't know about you, but this feels like a bug in the gps processing. Even a cold fix without ephemeral data should not take more than 30 minutes at the outer bounds.

It might take longer if the signal quality is low.

Alamanc data from a single satellite is ~12.5minutes (presumably it can be downloaded in parallel, depends on the software in the chipset), then ephemeris data for a given satellite is repeated every 30s.

If the signal is poor at any step, it will no doubt take longer.

Regarding the faulty nvd_data, if you tell the chipset that it should expect a certain set of satellites, I presume it must eventually decide that the intput data were wrong and start searching again (similar would happen if you started the chipset after a break of a week, etc., unless gpsdriver looks at the time that the last data were saved and it decides whether they will still be "in date").

It would be nice to see the source code....

By the way, I have tried putting a good quality spectrum analyser on the GPS band to check for interference - just in case. It was completely clear.

Also, the signals are good and strong looking at the N810s on my window sill.

In general there's a difference between signal strength as returned by the NMEA data (also known as AMUs - Arbitrary Mystery Units, see http://gauss.gge.unb.ca/papers.pdf/SNR.memo.pdf) and the signal quality.

I don't know whether you are working out the quality of the signal with your signal analyser (e.g. by processing the data using something like this: http://www.l3nav.com/gps_toolbox.htm)?

I was just looking for any obvious stuff above the noise floor with a preamp on, not demodulating the signal. This was just in case we were inadvertently jamming ourselves :-)

I have been trying out the fix on all five N810s this afternoon and it has worked every time. Sometimes it can take a couple of minutes to lock, but at least they always lock now.

The lesson seems to be that you should make a backup copy of nvd_data taken from a period when you know it had locked on okay. Then, never, ever, turn off your N810 or let the battery go flat if it is not locked on. I am 90% certain that this what caused both of mine to go bad as I reckon that they were turned off indoors. So, always turn off your N810 when you're standing by a window or outdoors. Oh, and it's probably a good idea to make sure that gpsd is turned off before pressing the off button.

Paxton

P.S. Thanks for the pointer to the Matlab toolbox. I'd like to find time to have a go one day.

The wonders of black box components! :S

Np, me too, what are you using to capture the raw signal?

if you want a raw signal, you can query the ethernet port on of the more modern analysers (e.g. R&S FSU, FSQ or one of the Agilents). That will give you I/Q directly. I prefer the FSQ as you can choose the sample rate and the resampler in the instrument does it all for you. I use the Matlab instrument control toolbox, but it's a bit messy to set up the first time. It might be easier with Labview.

I've been trying the N810s out in central London today. I managed to get a fix on every gpsd restart, although sometimes it took a couple of minutes. At least it works every time now - yey! - you just need patience.