GPS soft baseband processing

[quatrox at gmail.com]

On 2008-09-13, Joerg Reisenweber <joerg at openmoko.org> wrote:
> Am So  7. September 2008 schrieb quatrox at gmail.com:
>> On 2008-09-07, Al Johnson <openmoko at mazikeen.demon.co.uk> wrote:
>> > On Sunday 07 September 2008, Andy Green wrote:
>> >> Somebody in the thread at some point said:
>> >> | On 9/7/08, Flemming Richter Mikkelsen <quatrox at gmail.com> wrote:
>> >> |>  "Our automatic geotagging technology solves these
>> >> |>  problems. A fragment of RAW GPS RF data is captured
>> >> |>  in 0.2 seconds and then stored alongside the digital image
>> >> |>  and the processing is performed later on a PC. Because
>> >> |>  we separate the GPS capture from the GPS processing,
>> >> |>  our technology maintains the traditional "instamatic" user
>> >> |>  experience we are all used to and then pulls consumers
>> >> |>  to experience a far more interesting way to search and
>> >> |>  share photographs later."
>> >> |
>> >> | Interesting.
>> >> | Around 1995 I  used about the same technique to sync spectrometer
>> >> | data to an external authoritative timesource ( in that case the DCF77
>> >> | signal encoded into 10ms dataframes ) time recovery could achieve
>> >> | better than 100us
>> >>
>> >> It sounds like a fine technique, but it pushes the whole issue into
>> >> license of the host CPU software then.  They don't seem to mention if
>> >> they provide it or it is liberally licensed, etc.
>> >
>> > There appear to be several open source GPS projects around, including a
> few
>> > that do correlation in software.
>> >
>> >
> http://sourceforge.net/project/showfiles.php?group_id=91538&package_id=271229
>> > http://gnuradio.org/trac/wiki/GlobalPositioningSystem
>> > http://gnuradio.org/trac/wiki/GlobalNavigationSatelliteSystem
>>
>> The article I read mentioned all the details of what the chip does and
>> what is left for software. IIRC, it is basically an FFT and a reverse
>> transfer function which is needed. There is free implementations of
>> FFT. It looks like the only problem is that calculating everything each
>> second might take much CPU, but then again, we can choose to only
>> calculate every n'th  second or so.
>
> GPS chip has xxKunits of hardware correllators. I don't see benefit of
> moving
> all this to main CPU.
> You might take a photo in 1 second, where you might be able to caculate
> later
> on where the photo has been taken.
> But IIGTR you trade that in for ability to get a fast realtime first fix.
I think when we got a better CPU, it might be an option. But I agree
that there is pros and cons