[Date Prev][Date Next][Thread Prev][Thread Next] - [Date Index][Thread Index][Author Index]

An all software HF/weak signal modem

Yes Virginia, I am alive and well.

I have just gotten working a software implementation of
Mil. Std. 188-110-A1 modem using Soundblaster cards
and linux machines.  You must have a linux sound
system driver (using /dev/dsp) in order for this
to work.  It works in all modes, all speeds. 
(75 bps-4800 bps, short and long interleavers),
it has a mode that is not in the mil-std that is
suitable for AX.25 and is an emulation but not a copy
of the Rohde & Schwarz modifications to the Mil
Std. and used in the their products.  They
use R&S-HDLC, I will use HDLC and the AX.25
system in linux.  This is NOT compatible with their
stuff because I don't want to fool with their proprietary
stuff but uses some of their approaches.

I will unleash it at the ARRL/TAPR DCC:


and there will be a paper in the proceedings with all of
the hairy gory details.

It will be distributed with a Gnu P.L.   It is based on
new work done by me extending work done by Nikias,
et. al. on using the EM algorithm to do simultaneous
channel parameter estimation and data estimation
in a near optimal fashion.  This escapes the Harris
patent since it is not an equalizer at all.  It makes
no attempt to equalize.  For those unfamiliar with the
EM algorithm, I will give a good bibliography in the
proceedings paper.  The EM algorithm was invented
and proven to work at the place I work in the mid
60's but is still drifting into widespread usage even
after Bell Labs made such a big fuss over it in
Rabiner's shop.

The last trick after getting symbol probabilities out of
the channel in a great fashion,  is decoding.  The
code makes extensive use of these probabilities and
the signal structure (8 psk, Grey coded, etc.) to
estimate soft decisions on the bits which are then
permuted back to the original order (deinterleaved)
and fed into the ka9q decoder (thanks Phil!).

It easily runs real time on a 100 Mhz linux pentium.
On my dec alpha, running on a file, it can demod
203 second transmission in 9 seconds.

I will support anyone who wishes to port it to Windoze
and I have the Visual development tools to help, I
just don't want to learn directX, at this time (though
my kids do).  I will start giving it away on the TAPR
site after I finish three things:

(1) Cleaning up my mess so that others can read it.
(2) Fixing the performance of the "cut-in" demanded by the standard
(3) Making an implementation improvement in the correlation for
     known symbols and their use in frequency offset calculations.

See you in Phoenix.


Via the amsat-bb mailing list at AMSAT.ORG courtesy of AMSAT-NA.
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org