Return to KLUBNL.PL main page

rsgb_lf_group
[Top] [All Lists]

Re: LF: VO1NA Opera-32

To: <[email protected]>
Subject: Re: LF: VO1NA Opera-32
From: "Graham" <[email protected]>
Date: Sun, 23 Jun 2013 13:35:46 +0100
Importance: Normal
In-reply-to: <A8549327BF0C4BA098AB5F639CA8CC85@White>
References: <67BD952BF77F49FD826969BF2E7CC824@White> <B42304EC14DE45B28F559A159F5E4047@AGB> <A8549327BF0C4BA098AB5F639CA8CC85@White>
Reply-to: [email protected]
Sender: [email protected]
Interesting points Marcus ,
 
Assuming the  Tx station  declared  the  make up of the  call  , the  system could be  made to   work ?
 
As for the  differing  decode  dB levels  ,  there  is 'no'   actual  cliff , the  minimum   Op decode  level  is set  with  regard  to a  number  of  factors ,  ie  false  decodes  , lowering the  bar  may  result in more  'real' decodes  , but  with  a  increase  in false ones , at the  moment , it seems  to  be  'about right'  ? 
 
Noting  , that  Opera  was  designed to  accommodate  a  hardware  limitation, ie  to  replace a cw  key  , where  as the  wspr  system  conversely , dictates  hardware requirements , ie stability , remote sync time , vfo  control .
 
The  differing  modulation  systems  , do have  numeric differences ,  wspr  with  4 fsk  , will  have a  inherent  gain over  1 tone  OOK ,  OOK also  loosing   50% of the  'energy'  -3dB  , on paper  this  dose give an advantage to  4fsk  ,  but  over a  real  path , the  Op system , with  dedicated   data  processing  and  fec  , linked  with the  DSP  engines  design , that  can  maximise 'flutter' and reject  noise  , on air,   results  with similar  carrier  levels  are  very  similar  and  in high  noise , the  non linear routines  in the  dsp  are  able to  maintain performance  under quite  'extreme' conditions .  OP is a  averaging  system and  differs  significantly  from  conventional data  methods
 
The  only  , point that  should  be  stressed  , is , that  the  Op software  is  decoding  data , with  no  pre information   , the  call  sign or in the  case of the  QSO  mode , 15  chrs  plain text  , the  correlation  system is  detecting  with  a system pre-load , making the  decision as to  the  authenticity of the  detection
 
But in saying that , in keeping  with  LF  , an authenticated  detection  is  an improvement !
 
73 -G..
 
 
 

 
Sent: Saturday, June 22, 2013 9:24 PM
Subject: Re: LF: VO1NA Opera-32

Hi Graham,
 
good question... Edgar J Twining from Tasmania asked the same a while ago. You'll find the respective emails posted beneath. 
 
Best 73,
Markus (DF6NM)

From: Graham
Sent: Saturday, June 22, 2013 9:48 PM
Subject: Re: LF: VO1NA Opera-32

Marcus
 
Q   what  advantage could you  have , using  this to  detect  wspr-15  ? that  has a  fixed  pattern  and a  known start time ?
 
G..

Sent: Thursday, February 21, 2013 1:10 AM
To: edgar
Subject: Re: "Deep search analysis"

Edgar,
 
yes, at least in principle. I had actually looked into WSPR and QRSS correlation before turning to Opera. With Opera, it looks like I can detect up to 10 to 12 dB deeper than the standard decoder. But I think there is simply less to gain with WSPR.
 
The first step in my procedure is searching for a carrier. If Opera is transmitted phase-coherently, it's just classic AM, with 50 % of the transmitted energy contained in a central spectral line. This is relatively easy to detect eg. with a long 0.47 mHz FFT. On the other hand, WSPR shares it's "carrier" power among four equispaced lines, which lets them stand out less above the noise.
 
My second step then uses the detected carrier as a pilot to correct for small amounts of drift and propagational phase variations. This could probably be done equally well with a coherent sum of four carriers as long as their spacing is accurate. The third step, the actual correlation against a number of precalculated list of templates, should also be comparable for the two modes.
 
The main difference is however that Joe's WSPR decoder seems to be much more efficient in the first place than Jose's bit-power based Opera decoder. For same message time, the difference is said to be about 6 to 7 dB. So there is simply less to gain by using all the available signal energy in a correlation process.
 
Another difficulty is that WSPR actually transmits more information than just the callsign. So for preparing the templates, you would have to include locator and power as a priori information.
 
I have also looked at integrating subsequent repeats of WSPR. It is fairly straightforward to stack them incoherently, in essence like averaging the brightness of successive spectrograms. However this gains only 1.5 dB per doubling of aquisition time. Coherent superposition is more attractive (3 dB for double time), but requires extremely stable signals over long periods of time, and a milliHz search for the exact phase shift from one repeat to the next.
 
Best 73,
Markus (DF6NM)
 
----- Original Message -----
From: edgar
Sent: Wednesday, February 20, 2013 12:49 AM
Subject: "Deep search analysis"

Hi Markus,
 
    Can you "deep search analysis" that you applied to OPERA signals be usefully applied to WSPR signals?

Regards, Edgar.

 
<Prev in Thread] Current Thread [Next in Thread>