Re: [Discuss-gnuradio] Another question about GPS

[Matt Ettus]

> But if you don't know anything, in my understanding, you must:
>
> 1. Try every (32) PRN code, since you don't know what satellites are
> over your local horizon.

Yes

> 2. Try multiple frequencies in order to compensate doppler deviation of
> unknown satellites respect to your unknown position.

Yes

> Since PRN codes are repeated every millisecond and they are 1023 bits
> long and 32 PRN codes in total, suppossing a non parallel correlator you
> would need about 1-2 seconds per PRN code, 32-64 seconds in total. But
> you must do that searching in several frequencies, because the doppler
> thing.

Yes.

> So, my questions are:
>
> a) Are my appreciations basically right?.

Yes.  it is a very big search space.

> b) What is the maximum doppler offset you could possibly expect?.

+/- 5kHz from satellite motion.  You usually search +/- 10kHz if you don't know
how bad your own clock accuracy is.

> c) What could be the frequency "step" in that interval?.

If you search for 1ms at a time, you can do 500 Hz steps.  If you search for 10
ms at a time to find weaker satellites, you need to do finer steps.

>  From b and c, and the 32-64 seconds figure for full PRN code search, we
> could get the initialization time for a GPS "full dead cold start". An
> intuitive first aproximation, brute force search, could be "hours" :-p

It's not that bad.  But if you have a limited number of correlators and you
don't use fft methods it can be as long as 15 minutes if you are unlucky.

Matt