[pskmail] Re: Wanted: Pskmail technical docs/specs
- From: "Jörg-Michael Hasemann" <michael.hasemann@xxxxxxx>
- To: pskmail@xxxxxxxxxxxxx
- Date: Mon, 09 Jan 2012 19:57:15 +0100
Hi John,
many thanks for your extensive and really helpful answer. It will surely help
me ahead.
I wonder how the mode capability negotiation mechanism works. It uses mode
table (are those ones fixed or can they be arbitrarily filled with supported
modes) but how does the selection algorithm work?
Many thanks,73
Michael DB5MH
www.diskoisland.de
-------- Original-Nachricht --------
> Datum: Mon, 9 Jan 2012 10:54:07 +1100
> Von: John Douyere <vk2eta@xxxxxxxxx>
> An: pskmail@xxxxxxxxxxxxx
> Betreff: [pskmail] Re: Wanted: Pskmail technical docs/specs
> Hello Michael,
>
> 1. On the PSK-Robust modes:
>
> I will give you a couple of web sites that explain the design in
> general and a summary of the specs.
>
> http://www.w1hkj.com/FldigiHelp-3.21/Modes/index.htm
>
> Since the design of the convolutional coding, interleaver and
> character set is very similar (or identical for some parts) to the
> MFSK design below is link on that mode's design. If you go to the
> "Forward Error Correction" section you will get the essential.
>
> http://www.dxzone.com/cgi-bin/dir/jump2.cgi?ID=6218
>
> The PSK-R modes use both a convolutional encoder and an interleaver.
> The rate is 1/2 meaning we send two bits over the air for each bit of
> data.
>
> The convolutional encoder, based on the NASA design (parameters are K
> = 7 and polynoms in hexadecimal 6D and 4F). This is identical to most
> digital modes using such an encoder in Fldigi.
>
> The interleaver is a matrix (square) form as designed in MFSK but the
> length (i.e. the number of concatenated matrices) varies with the
> mode.
>
> The objective being to retain a constant time delay between the bits
> so that the performance under noise bursts (e.g. lightning crashes) be
> similar. The number of square matrixes of 4 bits is 40 for PSK125R, 80
> for PSK250R and 160 for PSK500R).
>
> Finally, the character encoding chosen was that of MFSK instead of the
> PSK one as it delivers around a 10% improvement in speed.
>
> Since the underlying modulation is PSK, we get one bit per symbol
> transmitted and therefore we have to transmit the two bits out of the
> encoder/interleaver in series, meaning that on the receiving side we
> don't know if a given bit is the first or second bit from the encoder.
>
> Therefore we decode two streams in parallel, with one delayed by one
> bit, then we choose the stream with the best decoding metric as
> output.
>
> In the decoder we also use soft decoding where the angular deviation
> is taken into account for the quality of the logical one or zero. The
> convolutional decoder uses that quality information to make it's
> decision for a given decoded bit (one or zero). In gaussian white
> noise tests, this adds about one dB of extra robustness.
>
> The combination of an interleaver with the convolutional encoder
> provides the major contribution to the extra robustness of the mode
> when compared to a standard PSK mode, to the point that in my tests,
> as well as observed in the field I get better results with a PSK-R
> mode than with the next slower PSK mode while the data rate is the
> same (meaning for example that PSK250R is more robust than PSK125 in
> my experience).
>
> Since PSK250 is 3dB less sensitive than PSK125, this means that the
> coding gain has to be more than 3dB to make the above statement true
> (in white noise).
>
> In real life noise occurs more in bursts and the advantage of the
> concatenated convolutional coding tends to be even better.
>
> Nevertheless I don't know of a Turbo code or LDPC encoding scheme used
> in amateur modes which in theory should give us another 2.25 dB or so
> of extra coding gain and bring us close to the theoretical limit of
> channel capacity (Shannon limit).
>
> Interesting stuff.
>
>
> 2. The ARQ layer:
>
> The specs are contained in the client or server's manual in the appendix B
>
> http://pskmail.org/downloads/PSKmail_server-1.0.34.pdf
>
> 3. The application layer:
>
> Explained in the server's manual (link above) section "Server Commands".
>
>
>
> Hope this helps. Don't hesitate to ask.
>
> 73, John (VK2ETA)
>
>
> On Mon, Jan 9, 2012 at 8:24 AM, "Jörg-Michael Hasemann"
> <michael.hasemann@xxxxxxx> wrote:
> > Hi,
> >
> > I am very interested in PSKMail and looking for technical specs
> regarding the PSK robust modes (e.g. the FEC-mechanisms employed), the
> protocol
> data unit formats, the arq-mechanism and the PSKmail-application protocol in
> general.
> >
> > All I found in the internet was source code that implements these
> methods. What I am looking for are docs/specs about above.
> >
> > Any pointer is highly welcome.
> >
> > Many thanks, best wishes, and 73
> >
> > Michael DB5MH
> >
> >
> >
> > -------- Original-Nachricht --------
> >> Datum: Sun, 8 Jan 2012 20:24:02 +0100
> >> Von: "Gunnar Bulukin" <gunnar@xxxxxxxxxx>
> >> An: pskmail@xxxxxxxxxxxxx
> >> Betreff: [pskmail] Nice!!!
> >
> >>
> >>
> >>
> >> This message has been scanned by Avast for Microsoft Outlook
> >> P Please consider your environmental responsibility. Before printing
> this
> >> e-mail ask yourself:- "Do I need a hard copy?"
> >>
> >
>
Other related posts:
