[pskmail] Re: Other digital modes for PSKMail
- From: Rein Couperus <rein@xxxxxxxxxxxx>
- To: pskmail@xxxxxxxxxxxxx
- Date: Sun, 11 Nov 2007 10:18:10 +0100
One of my next projects will be memory arq (you will find some code is already
present in
the sources...).
One of the problems with memory arq and psk is the varicode, which generates
certain effects
during qrm/qrn. Often the number of characters in the block will be wrong. That
is why
memory arq at bit level is not effective.
What I have been testing is memory arq at character level. This makes use of
certain redundancies
in the header/CRC information. What you do is store a copy of the wrong packet.
When you receive this block again, and it is again wrong you try to reconstruct
the packet by combining
parts from both by looking at the start and the beginning of the blocks. That
also works
for punctured blocks with holes in the middle. This still needs a lot of work
but it is on the agenda....
73,
Rein Pa0R
> -----Ursprüngliche Nachricht-----
> Von: pskmail@xxxxxxxxxxxxx
> Gesendet: 10.11.07 18:36:59
> An: pskmail@xxxxxxxxxxxxx
> Betreff: [pskmail] Re: Other digital modes for PSKMail
>
> Hi Vojtìch,
>
> I thanks for the reply and the information. My short distance tests (a
> few KMs) support your comments between Olivia and MFSK16, with MFSK16
> more reliable (at least in the implementation of Fldigi). But I need
> to do some real life tests at several hundred KMs to be able to
> comment.
>
> Also I found DominoEx having difficulties for some reason when the
> theory would put this mode in front of MFSK in terms of robustness.
> Again, more tests are required.
>
> Regarding RSID, with PSKMail the frequency is known in advance, so I
> suppose it is less critical than with FlARQ where linkups can be
> ad-hoc, unless the frequency stability is an issue during the
> exchange. On that subject I noticed during my short distance tests
> using MFSK16, that at every exchange the receiving Fldigi would take
> about 3 to 5 seconds to show a full "green bar" of the link quality. I
> suspect that could be related to the frequency re-alignment.
>
> When I use PSK, the green bar goes very rapidly from low to high at
> the beginning of the reception of a frame.
>
> PSK with FEC may very well be a plus there. As Rein mentioned before,
> the faster the transmission rate the less chances for a QRN hit. But I
> have experienced several time lighting strikes noises that so close to
> each other than there is no chance for a single packet to go through
> complete or that so few packets go through that the synchronization
> between server and client is impossible.
>
> It may bring the best of both worlds: redundancy and small
> bandwidth/performance of PSK.
>
> I hope Dave can implement in Fldigi.
>
> 73s,
>
> John
>
>
> On Nov 9, 2007 11:26 PM, Vojtìch Bubník <bubnikv@xxxxxxxxx> wrote:
> > Hi John.
> >
> > > I would like to experiment with other modes. Logic tells me that a mode
> > > with
> > > longitudinal FEC of some kind should help since the part affected by the
> > > strike could be re-built provided it is small enough. What would you
> > > propose
> > > I try first? Olivia or MFSK16 or others? Speed can be reduced, but I would
> > > like to keep a reasonnable bandwidth (500Hz if possible, and not more than
> > > 1000Hz).
> >
> > I would recommend MFSK16 with a RSID sent in front of connect request /
> > connect response packets. This way both sides would know their exact
> > frequencies. Both sides would need to have stable enough VFOs though. Oh,
> > pitty, RSID is currently implemented only in MultiPSK, but Patrick is
> > willing to give away source code to the developers.
> >
> > Olivia coding is less effective than MFSK16, but on the other side it has
> > strong correlation, by which the signal may be localized in frequency.
> >
> > How about PSK63F or PSK125F? PSK63F is a mode derived from PSK63, but uses
> > MFSK16 FEC and coding. I will release PSK63F support in PocketDigi soon
> > hopefully, Dave may port it to flDigi.
> >
> > 73, Vojtech OK1IAK
> >
> >
>
--
http://pa0r.blogspirit.com
Other related posts:
