Re: [gpsd-dev] PPS over USB

[Ed W]

...Very hostile answer Gary? What did I say to tick you off?


> > Just to be clear - the above is a waaay more powerful board than most
> > routers, it's i586 compatible (rather than ARM).  However, it is
> > probably more expensive than an old Linksys router, and it's
> > switching capacity is almost certainly lower than a cutting edge
> > Linksys job.
> And that is exactly the problem.  They want a common dumb consumer
> platform becasue they are looking to develop teaks for dumb consumer
> platorms.

Still not quite clear what you are saying here?  First you say you can't 
do diy, then you say that you want to tweak off the shelf stuff?  At 
least some interpretations of that are incompatible? Are you saying that 
you cannot narrow the platform hardware choices?

The front page of the thumbgps platform suggests you are building your 
own 'wrt clone?

Note although you are ticking me off for *asking* if DIY is an option, 
your mailing list is full of technical chatter - I think you are quite 
harsh to an outside simply trying to offer some ideas when it's really 
not obvious what the parameters are.  I will stand corrected though...


> > The bit I was trying to square was that the GPIOs don't currently
> > have a linux driver which is interrupt triggered.
>
> Yup, and the bufferbloat project noted that and explictly decided that
> to go down that road was to go Yak Shearing.  If such a driver should
> appear that might be usefull in a future project.

Don't understand?  Isn't software on the table?  I thought it was only 
diy hardware which was out of contention?

I already managed to get the basic GPIO changes into kernel 3.2 to 
support this.  Either I or someone else needs to beat me to take the 
code from the OLPC modules and turn it into a generic module.  For sure 
it's systems programming, but probably around 100 lines of code or so.

The last 6 months have had me out of action, but I might get a chance to 
look further into this shortly

I do agree it's not perfect hardware though - hence my questions on what 
your requirements were.  Had this been a useful solution I might have 
been inclined to try and help get such a module developed sooner



> > > That said, I would personnally love to find a USB 2.0 equivent to
> > > the USB1.1 PL2303.
> > Just clicking on the first hit in google takes me to here.  Is this
> > what you want?
> >       http://www.tronisoft.com/prod.php?id=2327
> Nope. That is just a USB to RS-232.  The problem is finding a
> matching cheap RS-232 GPS, with PPS and preinstalled RS-232 connector.
>
> I am looking for an integrated product requiring zero hardware skills to
> use.

You are shifting the goals on me...

You appeared to ask for a chipset which was USB2, I think I found you 
one?  Yes, it's not a complete solution, but I wasn't clear what you 
were asking for

That said, I just pushed buy to see what they were like.  Assuming I 
wire a spare Garmin 18 up to one, is there any special way I need to 
configure GPSD to try and observe decreased jitter?  At least it's an 
experiment which will show if there is any mileage in pursuing USB2 options?


> > It's not bad, but I think the issue is still that the USB bus is non
> > deterministic, so it might still have a good bunch of jitter around
> > that.
> Actually, it is VERY deterministic.  USB 1.1 is polled 1024 times a
> second, and that is exactly what we see in the results.

I thought we were talking USB2 not 1.1?

In any case my misunderstanding was directed that the polling had a lot 
of jitter on it.  Also I was under the misapprehension that bulk 
transfers and the realtime class could further skew that timing - I 
concede I haven't looked into the details though

Some full PC motherboards appear to have more than one USB bus.  No idea 
on cheap routers (would guess not), but it would be worth allocating 
peripherals to cause least interference if it's an option...



> > Anyway, I got stalled, but my goal was quite similar.  I wanted to
> > try and offer a cheap solution to Stratum 1 servers.  The idea of the
> > Alix plus a cheapo serial GPS seemed very attractive.  They can be
> > powered via ethernet which potentially makes them easy to put into a
> > hobby box and put the whole box on the roof...
> And as a pre-packaged commmercial solution it might be very interesting
> to many people.  Just not the bufferbloat folks.

I would be happy to donate a bunch of these if there was interest.  I 
have a vague interest in trying to flood the internet with decently 
accurate Stratum 1 time servers that you can use directly - the current 
stratum 1 servers seem to be .... not that accurate ... and stratum 2 
cannot be more accurate than the stratum 1.


Note that the Alix cannot achieve greatness off the shelf.  The GPIO 
interrupts appear to have about 30 microsecond lag to be read.  Plus a 
relatively slow single core processor and poor default clock stability 
means we are unlikely to be vastly under millisec accuracy.  However, 
that seems to be a good deal better than the average Stratum 1 today

Ed W