[aprssig] I'm brainstorming, so don't beat me up!

[Jon Adams]

Interesting ideas, each of them. While sound-card based generation and
reception has become blindingly simple over the past few years, that doesn't
make doing a simple device or a non-PC based application any easier to do. 

Some basic things that you want to do for a mobile data system:

* Pick a combination of modulation, coding, spreading, protocol methods that
works for the expected propagation channel. Note that some of the more
interesting desires are mobility at potentially a few hundred kmph, so it
has to deal with multipath (Rician and Rayleigh) or flutter, rapid retries,
relatively low latency, some expectation of quality of service...
* What is the reliable expectation of average, delivered data rate in a
moderately busy channel? 1200 bits? 4800, 9600? (Right now we're working
with MAYBE 100-200 bps.)
* Do you want to be able to implement the system in a simple
hardware/software manner? 8 bit micro, 16-bit DSP, software signal
processing, etc? Do you want devices that are transmit only and blindingly
simple? Or blindingly simple transceivers? 

The list goes on. 

Now, once you've written down on the desirements, look at the practical
reality. If you pick a system too complicated to implement in hardware, so
that it takes either a PC and sound card to do or a $10M investment in
silicon designers to create a chip, you've probably killed the mobile,
simple to implement part of the quest. 

Use what's already out there. Shop around for silicon based solutions that
do the job, and augment those in hardware or software to customize as
required. Lots of xFSK transcievers out there (simple but very poor
performance); xMSK transceivers which are well behaved in a voice radio, and
are moderately easy to implement; xPSK transceivers are much better in
performance but not so common (look at IEEE802.15.4 protocol transceivers as
a basis there, especially the 900MHz ones); OFDM chips which are the basis
of wireline, xDSL and cable modems.

Once you've looked at all the possibilities, it seems to me that the nicest
opportunity here is the 15.4 chips. The biggest drawback being payload size
(just over 100 bytes per packet), these transceivers are cheap,
self-contained, and in fact if you can use them at their 900 or 2400 MHz
frequencies, ready to go out of the box. Still needs networking software, as
the radios themselves generate frames that have only a source and
destination address, and no routing info  within. In the 2400MHz band, the
over the air data rates are 250kbps, while currently in the 900MHz band its
40kbps with upgrades coming along to scale it to more like 250kbps in the
next year or so. The standard is built for high multipath environments, but
wasn't particularly envisioned for 10's to 100's of km range over a single
hop, so we might have to futz with code sequences or retry periods to patch
some of that. Channel widths are 2MHz or 5MHz, so we're not talking high
efficiency from a bits/Hz PoV, but we are talking cheap and available. 

Check out Freescale, Chipcon, Atmel, ZMD, others for chips for this space.
Those could really work out to be the inexpensive transceivers of the future
for us hams, and could find their way into applications that are very simple
and standalone, and battery operated. Prices in 10k manufacturing quantities
for a MFJ/Kantronics type vendor are in the $2.5-$4 range, and probably
available in onesies from Digikey or Mouser for double that?

Jon N7UV

Date: Thu, 07 Jul 2005 21:39:26 -0400
From: Wes Johnston <aprs at kd4rdb.com>
Subject: Re: [aprssig] I'm brainstorming, so don't beat me up!
To: TAPR APRS Mailing List <aprssig at lists.tapr.org>
Message-ID: <42CDD94E.2010406 at kd4rdb.com>
Content-Type: text/plain; charset=ISO-8859-1

What about q15x25? mix32 (or is it mix-w) supports it... agwpe is supposed
to Real Soon Now. q15x25 is 15 PSK carriers with an aggregate bit rate of
2500 baud on normal channels, and FEC. It's just a tweak processor intensive
to encode and decode.

Wes


Gerry Creager wrote:

> Actually, the 2-tone modem history pertains to telephone line 
> optimization for tone detection.  Preemphasis and deemphasis were 
> rampant, and tone detection was done with LC lumped constant filters.
> A single tone detector had a higher error rate.  Once you get to two 
> tone systems (remember RTTY?) one can start thinking about different 
> encoding and modulation schemes.  When one is thinking in terms of 
> bits/transition, multiple tones or constellations thereof start making 
> sense.
>
> I suspect we could create a high performance CW system using DSPs but 
> we're talking now about transitions per bit rather than bits per 
> transition.  For what we're talking about, does it make sense?
>
> gerry