FEC paper, implementation

[rob@sys3.pe1chl.ampr.org (Rob Janssen)]

Hi Phil,

I have received your paper and have read it a first time.
(it was no problem to fformat it on my Linux system)

It is very interesting indeed (as Geert Jan already said), although for
practical purposes I foresee more use on point-to-point links than on a
local network access.  We are now in the process of removing 1200 baud
AFSK access from the network nodes in the Netherlands, and that already
causes so much stir-up that I frankly cannot believe the amateur
community will leave plain AX.25 over cheap modems and transceivers
anytime soon...

However, on point-to-point network links the situation is different.
Such links can be replaced one by one, and experiments can be done on
links that have problems with interference.  Here I think FEC could have
a place, when the errors we are experiencing are likely to be cured by
it.

I think a nice platform to do these experiments would be the DSP4.  It
is a Finnish DSP56001 card (100*160mm euro) that includes a Motorola
56001 running at 27MHz, 32K*24 RAM, 32K*8 (E)EPROM, a stereo CODEC that
can do up to 48Ksamples/s, a 19k2 bps RS232 interface and some generic
I/O lines that can be used for transmitter control etc.

A convenient monitor program is available that initializes the devices
and allows for downloading programs via the RS232 interface or booting
software from the (E)EPROM.  Some modem software is also available.
Full documentation can be found on jeeves.hut.fi in /pub/alefnull/doc.zip
(Postscript format)

While this card is probably not capable of doing 56Kbit/s, it could be
used to make a complete FEC system at some lower speed, fed using some
KISS-like protocol and directly interfacing to the modem hardware.
Thus, it could implement some of the more advanced features you
describe, because there is direct access to the A/D converters.
It also would fit in nicely with the link hardware that we are currently
using in the network, and that is not so likely to be replaced by
56Kbit/s hardware soon, because of the higher complexity, cost, and
bandwidth.

I have one of these cards (built and working), and there are a few
others in the Netherlands as well.  New boards are just being produced,
and should be available soon, I think.  I'm sure the makers of the card
are interested in this kind of use.

My card has been sitting on the workbench awaiting some real use (I have
done some experiments with filters and signal generation), but the
async interface (and thus the use of KISS) has kept me from doing more
on it.  I think KISS is not suitable for good implementation of AX.25
connections.  I have written a document that details an extended version
of KISS including report-back of actual frame transmission and CRC check
of the data on the RS232 line, but that has not yet been implemented in
the DSP4 monitor program.

Although I can't promise to do any work on FEC on this platform (I have
too many other unfinished projects), maybe you or someone else on the
list is interested in taking this up?

73,
Rob

(I have cut down the CC list a bit.  this was the original distribution
list of my modulo-128 paper)

--
+------------------------------------+--------------------------------------+
| Rob Janssen         rob@knoware.nl | AMPRnet:   rob@pe1chl.ampr.org       |
| e-mail: pe1chl@wab-tis.rabobank.nl | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU |
+------------------------------------+--------------------------------------+