Re: [Thumbgps-devel] Build vs. modify vs. what should we be doing anyway?

["Eric S. Raymond" <esr@thyrsus.com>]

Patrick Maupin <pmaupin@gmail.com>:
> On Mon, Mar 12, 2012 at 4:03 PM, Eric S. Raymond <esr@thyrsus.com> wrote:
> 
> > I'm still skeptical about custom builds for production - means one of us
> > would have to do the soldering, keep inventory, ship the results.  I'd rather
> > develop a design and talk SparkFun or Dangerous Prototypes into fabbing it.
> 
> But if you work with somebody like this or like seedstudio, they will
> order parts for you and assemble according to your instructions.  If
> you simply view the dongle as a part to be ordered and the assembly to
> be "take it apart, add this blue wire, put it back together" you will
> realize that the same process could apply, at least for a company that
> does more than basic board assembly.

Good point.  Then, though, we're dealing with the risk that our supplier 
will EOL the product we're blue-wiring.  I can tell you from experience
that consumer-grade GPSes like these have ridiculously short lifetimes.

A friend (who may join this list shortly) once explained to me that
this is a result of patent lawsuit risk. The big Taiwanese electronics
companies spin up small shell companies to own the product designs, rent time
on their fab lines, sell a shitload of units from a couple of big-batch
runs, then fold before the lawsuits can land leaving the parent company
legally in the clear.
 
> > [Test system] sounds like a larger project than just reverse-engineering the dongle,
> > though.  Do you have the capability, personally, to design and build such
> > a thing?
> 
> This is more what I do on a daily basis.  Use a bunch of readily
> available off-the-shelf thingies to make a few units.  In this case, I
> would buy a third party FPGA board, make a couple of smaller PCBs to
> talk to it with a socket for an OCXO, maybe a connector to accept a
> clock from a rubidium time source, headers for some USB modules and
> some GPS modules and do some FPGA coding to allow reliable high
> resolution timestamps.  There are a couple of levels of this.  The
> first is just to do some one-off futzing around looking at various
> devices.  The second is to get serious.  You could probably build 3
> identical units with really good resolution and stability (expensive
> OCXO) for under $2K total.  Maybe a lot less if you can find some good
> surplus stuff on ebay.

OK.  Hal Murray will probably want one of these.
 
> > And this makes complete sense.  Seems like a good way to start that doesn't
> > require excessive time commitment up front on your part.
> 
> In that case I'll start ordering and plotting out what I'm going to
> do, and will sign up for the project to start putting out information
> on the test setup for people to throw rocks at.

OK.  I'll keep working the net looking for products we can mod or
reverse-engineer.  Also trying to cultivate interest in building 
product for us at Sparkfun and similar outfits.
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>