[aprssig] Sub-second timeslots

[Scott Miller]

I announced on the OpenTracker list yesterday that I'd had some success
in improving the OpenTracker's time synchronization.  With my test
firmware, as long as GPRMC is transmitted by the GPS receiver at the top
of the second, synchronization seems to be within about 40 ms even
without a 1-pps signal.

Yesterday I had two trackers running in adjacent 1-second time slots
without interference, which was something it couldn't reliably do
before.  Today, I've made an experimental modification to select the top
or bottom of the second for the time slot, and I've got two trackers
sending without collisions in a 1-second span.

Here's an audio capture of the two trackers:  http://n1vg.net/2packets.wav

They both finish in about .9 seconds, with .12 seconds of dead air
between them.  They're running Base91 compressed mode (including
course/speed) with no comment and no path.  Obviously you're not going
to have time to digipeat packets on the same channel if you're running
half-second time slots, so the lack of a path doesn't make much difference.

This should be good for special events and such - you can get 10
trackers on a channel with a 5-second update cycle.  There's more
testing to be done (I'd like to fill several adjacent slots and do a
long-term test) but the results so far are encouraging, and it's working
with cheap HTs, not commercial mobiles with fast TX/RX switching.

It'd also be nice to establish some sort of actual specification for all
this.  I don't think anyone's ever made any effort to make sure that
time slot implementations are compatible between vendors.  This really
needs a clearly defined spec and some interoperability testing.

Scott
N1VG