Another false decode

73, Spiros/SV8CS

Sent: Saturday, January 10, 2015 5:37 PM
To: [email protected]
Subject: Re: LF: Opera improving

The new checking system has just worked. JA8SCD was reported by SV8CS
at nearly 10,000km on 136kHz whilst the JA was using a dummy load.
This is plainly a false decode, but the new software showed "Deep
Search ??" to indicate this should be viewed with caution.

Mike, G3XDV
===========

Power gain is  close  to  6  dB  , so  that's  a  times  four  over
the standard   Opera  data  decode limits

Any  numerically  processed  data   mode  is  based  solely on  Guess
work, just that  Guess   work  is   translated  by  decision  making
and   fuzzy logic in to a  reproduction  of the  original  message

Modes like  opera / ros / olivia  etc  use  algorithms  to  compress
and extract the  call , hence  opera  data  miss  decodes, 'have'   a
normal call  profile  , but  are  not  valid . simple   rtty  just
produce's  near random  data

In the  case  of  Marcus's  opds ,  your  are  presented  with  a set
of values ,  a little  deeper ,  these  will  of  been  constructed
from  a layer  of  fuzzy logic , as  are  the  dsp  decisions  in
opera  etc ,  in the  case  of Opera dynamic ,  the  outputs  from
the  data  decoder   and the  'search'  detector  are  assimilated by
a another  layer  of  decision making  logic

The  command  structure ,  then  decides  at what point  the  decoder
has failed  and  the  results  from the  detector  are  evaluated ,
the dynamic call  lists  are  maintained , via tx  web  beacon  and
1+1  quality check of  decoded  calls

Table  is  held  for  8 days  We did  have  3 days  as  the  original
cycle time ,  however ,  as   activity on  136/477  tends  to  follow
the  7 day week  pattern  , stations  using  pic/pi/U3  and  other  Tx
 systems ,  with no  web-beacon  would  need  to  be decoded  by  2
stations  using op32/op8  on there  first  pass, to  catch  the
second

The show go's on ;)

73-Graham
G0NBD





--------------------------------------------------
From: "Mike Dennison" <[email protected]>
Sent: Saturday, January 10, 2015 11:09 AM
To: <[email protected]>; <[email protected]>
Subject: LF: Opera improving

> There is a British saying "Don't throw the baby out with the
> bathwater". Amongst the criticism of false decodes (which I have
> been part of), we must not forget that Opera is now more sensitive -
> the equivalent of doubling everybody's power - which must be good.
> This has led to many more good DX decodes.
>
> Now, in response to suggestions from users, the writers have
> modified the program to flag up decodes that might be suspicious,
> which is what OPDS does (but in a different way). Used with caution,
> this will address most of the problems experienced with the launch
> version.
>
> The controversial PDF file has also been dropped from the zip file.
>
> Deep Search, originally developed by DF6NM, has always involved
> "guesswork", and therefore needs to be viewed with caution. In short
> it compares a list of known active stations (again Op and OPDS does
> this differently) with what might be described as 'partial decodes'
> and finds the best match. Anyone who has worked DX, operated in a
> contest or used QRSS will be familiar with that technique. Indeed
> that is how the human brain works all of the time - it matches what
> its sees with a huge look-up table of past experiences, which is why
> we see faces in clouds (a false decode). The key is to use our
> experience to work out what is true and what is false.
>
> I would suggest three more improvements:
>
> My understanding is that callsigns in the 'look-up table' are
> dropped after 8 hours. Reducing this period might help.
>
> For those who don't like Deep Search, there should be an 'off'
> switch, perhaps in the existing Deep Search menu. The 'on' swtch
> might come with a "use with caution" message.
>
> Used carefully, the new v1.5.6 should be a great improvement on the
> (pre Deep Search) version 1.5.4.
>
> Mike, G3XDV
> ===========
>



-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2015.0.5577 / Virus Database: 4257/8901 - Release Date:
01/10/15