Re: LF: Opera -Just a 2 mins carrier gives ghost detection.

["Graham" <g8fzk@g8fzk.fsnet.co.uk>]

Jay,

 

The whole world  of data   is based  on  probability   , under the  'hud  a  whole sequence  of  decisions  have been made , before the  data  is released ,  in this  case , Opera uses  data  from  other sources  to  double  check , either  multiple  spots  or the  time  stamp as the final  check

 

With a  one off  single  detection  ,  its down to  the  time  window ,  vk5zuc  posted this  in the  600m  group , which points to   5% error , that  could even be  a little  shy  of the  real  number.

 

looking  over the  last  24H ,  no  false  spots on 2200  and   2    dynamic  showing  on  600  and  2   Opera  from a  pool of  app  22  stations  , Both  systems  run  in parallel  making  2 per system  

 

To achieve  the  same  % error  rate  from a  correlation  system  ,  compared  to a  data  decode , with a  gain of  5/6 dB  is  quite a  achievement , sure  , more gain can be added  , but  so  is the  % error  rate increased

 

Which brings  me back to  my  first  posts , ' the  up grade  is  fit  for  purpose '

 

73-Graham

G0NBD

 

 

Ron, Graham and others.

Is there a way of estimating the probability that a detection is real or false.

Perhaps Bayes' theorem can help define true positives.

I propose the scientific probability of 95%. So there is only a 5% chance that it is false.

Do we need more certainty than that?

Andrew VK5ZUC

 

From: jrusgrove@comcast.net

Sent: Saturday, January 17, 2015 1:15 AM

To: rsgb_lf_group@blacksheep.org

Subject: Re: LF: Opera -Just a 2 mins carrier gives ghost detection.

Graham

 

Perhaps deep search is not appropriate for 'standard HAM kit' then.

 

With DF6NM's OPDS we have the correlation % and dB 'certainty' indicator, an accurate time indicator of the transmission and a highly accurate frequency readout to base our judgement on.

 

With Jose's software we have no such information to make a judgement. As you say 'everything is inside' so we are expected to take the software's decision as gospel! The fact that someone somewhere happens to also receive a spot (within a 2 - 5 minute window) is woefully insufficient for validation. 

 

Ask Jose to make the code available for inspection so we can assess the decision process.  

 

Jay W1VD  WD2XNS  WE2XGR/2

 

 

  

 

 ----- Original Message ----- 

From: Graham

To: rsgb_lf_group@blacksheep.org

Sent: Friday, January 16, 2015 5:18 PM

Subject: Re: LF: Opera -Just a 2 mins carrier gives ghost detection.

Jay,

 

The house style is everything is on the  inside ,  Ros/Opera  are  designed  to  achieve  the  best  performance  from   standard  HAM   kit ,    in the  case  of the  dynamic  option ,  high  stability will  reduce the  processing times ,  but the  final  s/n  is roughly  the same,

 

processing  algorithms are used  to  minimise the  level  of  false detections against s/n ,  the  spots that  are  wrongly  validated  and presented, can be  tested  against  other  users  , as  shown below ,  or   against the  data  in the  pak-map  data  base   

 

 two detections ,  same  call  same time  stamp  is taken as validation , 

 

G,

 

 

 

Hear  G8LCO and  others  decode  in  OP8 ,  21:53 Gw0EZY + Myself  21:54  ,   ie  + 1 min

21:53 477 LA3EQ de G0NBD Op8 Deep Search 488 mi -35 dB in Liverpool with 0.01w + Marconi-T

21:53 477 LA3EQ de GW0EZY Op8 Deep Search 541 mi -35 dB in Welshpool IO82ho with 0.01w + Marconi-T

21:52 477 LA3EQ de G8LCO Op8 513 mi -26 dB F:12% in Welwyn Herts with 0.01w + Marconi-T

Same  in this  sequence , but  G8HUH  is   false ,  No  other  decode  , system  stamps  with  '??'

21:28 477 SV8CS de SV8RV-2 Op8 6 mi +3 dB F:3% in Zakynthos (Zante) isl.GREECE with 0.1w + "T" Marconi

21:28 477 SV8CS de SP5XSB Op8 995 mi -7 dB F:14%_Ini in Raszyn nr Warsaw with 0.1w + "T" Marconi

21:25 477 G8HUH de LA3EQ Op8 Deep Search ?? -37 dB in Egersund, Norway

21:24 477 LA3EQ de SP5XSB Op8 Deep Search 728 mi -37 dB in Raszyn nr Warsaw with 0.01w + Marconi-T

 

 

--------------------------------------------------
From: <jrusgrove@comcast.net>
Sent: Wednesday, January 14, 2015 11:01 PM
To: <rsgb_lf_group@blacksheep.org>
Subject: Re: LF: Opera -Just a 2 mins carrier gives ghost detection.

>> More than  one  detection/decode  is  then  used  to  certify the  spot.
>
> Certify the spot? The fact that someone else somewhere received a spot means that your spot
> certified good? Really? No further corroboration necessary?
>
> With DF6NM's OPDS system there's the correlation % and dB 'certainty' indicator, an accurate time
> indicator of the transmission and a highly accurate frequency readout that are all used for
> corroboration.
>
> Jay W1VD  WD2XNS  WE2XGR/2
>
>
>
>
>
> ----- Original Message -----
> From: "Graham" <g8fzk@g8fzk.fsnet.co.uk>
> To: <rsgb_lf_group@blacksheep.org>
> Sent: Wednesday, January 14, 2015 4:17 PM
> Subject: Re: LF: Opera -Just a 2 mins carrier gives ghost detection.
>
>
>> Jan,
>>
>> As posted , any decode  in dynamic , will  only  show  calls  from the  list , this  is the  bill
>> for  the  extra  5/6 dB  gain ..
>> your spots  are  pre-fixed   with  ''??''   this  is  added  by  the web-linked  system and
>> denotes  a  unique  detection , false Opera  decodes have  a  call  sign  format  , with  random
>> characters.
>>
>> Deep Search ??
>>
>> More than  one  detection/decode  is  then  used  to  certify the  spot
>>
>> To  check  the  validity ,  you  can  use  the  psk-data  base , which holds  the  web-tx beacon
>> and  rx-spot time  stamped ,  Opera 8 Dynamic resolves  in 0>2  mins  , Op32Dy  2 >5 Mins  after
>> tx /end/decode
>>
>> Otherwise , the  system  provides  the  quality  check ...
>>
>> false ....
>> 19:31    477 SV8CS de NL9222 Op8 Deep Search ?? 1260 mi -35 dB in Mijdrecht with 0.1w + Marconi
>> "T"
>>
>> Ok ...inside  the 2<>5 Min  processing  time  for  op32 dy , G3XDV validating  the  tx  /rx  time
>>
>> 20:21    136 G8HUH de EA5DOM-RX Op32 Deep Search 889 mi -41 dB in Albir RX Site
>> 20:17    136 G8HUH de G3XDV Op32 124 mi -13 dB F:0% in Welwyn IO91VT
>>
>> 73-G,
>> --------------------------------------------------
>> From: "pa3abk" <pa3abk@kpnmail.nl>
>> Sent: Wednesday, January 14, 2015 8:38 PM
>> To: "blacksheep" <rsgb_lf_group@blacksheep.org>
>> Subject: LF: Opera -Just a 2 mins carrier gives ghost detection.
>>
>>> This eve I monitored MF Opera, however my WSPR timer was operational the same time giving a 2 min
>>> carrier at h10, h30 & h50.
>>> Sorry for the unwanted QRM, however to my surprise NL9222 detected following:
>>> 18:51    477 SV8CS de NL9222 Op8 Deep Search ?? 2027 km -35 dB in Mijdrecht with 0.1w + Marconi
>>> "T"
>>> 19:11    477 LA3EQ de NL9222 Op8 Deep Search ?? 693 km -37 dB in Mijdrecht with 0.01w + Marconi-T
>>> 19:31    477 SV8CS de NL9222 Op8 Deep Search ?? 2027 km -35 dB in Mijdrecht with 0.1w + Marconi
>>> "T"
>>>
>>> NL9222 is 42km from me and receives my sigs at +1dB
>>>
>>> No other stations detected these stations. I assume they were not QRV either.
>>> Maybe on LF?
>>>
>>> I have the feeling that Opera DS is acting more human than desired.
>>> It's the same when a CW human taker is decoding a plain language text in heavy QRM at QSA 0.
>>> You assume that certain patterns is forming a plain word which match the look_up table you have
>>> stored in your brain.
>>>
>>> Jan/pa3abk
>>>
>>> --
>>> pa3abk <-> dordrecht jo21it
>>> Sent by Philips STB-75 with T100
>>>
>>>
>>
>
>