Re: [gpsd-dev] PPS over USB

[Gary E. Miller]

Yo Ed!

On Fri, 04 May 2012 20:39:20 +0100
Ed W <address@hidden> wrote:

> Just curious but is the goal: accuracy, cheapness, off-the-shelf or 
> something else?

Eric has been involved in the 'thumbgps' project.  A component part of
the 'bufferbloat' project.

The goal is to setup about 100 home routers with accurate (1 milliSec)
local time sources to be used as data collecting routers to understand
bufferbloat better.

To meet that goal they want to source about 100 inexpensive USB/GPS for
the local time source.  Having them all identical should reduce any
timing errors introduced by subtle differences in models.  These are all
software guys and previous efforts to get them to solder a few wires
have been a failure.  Thus the desire to source a commercial solution.

Navisys has sold us samples of a proposed GR-601W which seem to be
suitable.  Unit cost should be under $40 and time accuracy better than 1
milliSec.  Unless we find a show stopper bug, or a better solution, real
soon, the project will likely go with the GR-601W soon.

More info on the main project is here:
        http://www.bufferbloat.net/projects/cerowrt/roadmap

More on the thumbgps project ehre:
        https://lists.bufferbloat.net/pipermail/thumbgps-devel/

> I admit I haven't tried it, but I sell loads of PL2303 adaptors that 
> just happen to dismantle very easily.  I don't see any theoretical 
> reason why you couldn't easily modify one to either put 5V power on a 
> serial pin, or just drill a hole in the case and poke wires directly 
> inside.  This would allow you to grab say an old Garmin 18 off ebay
> and use that?

Ugh.  The Garmin 18s are unsuitable for indoor use.  The uBlox6 in the
GR-601W works well in my basement.

> I got part way through the implementation of using the GPIO inputs as 
> PPS on the PC Engines Alix boards.  The goal was to enable a fairly 
> cheap machine that could connect pretty much directly to something
> like a Garmin 18 (or the rather nice looking Skytraq modules, some of
> which seem to offer seriously accurate PPS options) and give you an 
> inexpensive Stratum 1 box. (PC Engines boards are circa $110 for 
> basically an entire 500Mhz i586 including ram, processor, USB,
> mini-PCI and up to 3 network ports)

That would be fairly similar to what the bufferbloat project is
looking to do, except they want it also to work as a standard
router.

> So, just trying to understand if the goal is to be more off the shelf 
> than a DIY prolific bodge?  Or if you want an entire Stratum 1 box
> with microsecond accuracy, then perhaps the Alix route gets you the
> finished product with 2W consumption for decently low money?

Off the shelf, repeatable, interchangeable and simple enough for
programmers to use is the requirement for thumbgps.

That said, I would personnally love to find a USB 2.0 equivent to the
USB1.1 PL2303.  Doing so would increase the time resolution from about 1
milliSec to 125 microSec.  An 8x improvement.  Right now the GR-601W is
three orders of magnitude worse than local PPS over serial and 16 times
worse than being on GigE adjacent to an ntpd server with serial PPS.
Getting to 125 microSec would make the USB/GPS only 2x worse than being
adjacent to a serial PPS.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97701
        address@hidden  Tel:+1(541)382-8588

signature.asc
Description: PGP signature