Graham thinks that because there were no more false decodes in the previous
24 hours this one MUST be good.
Its not Just Good Alan ... Its 'Ruddy MARVELLOUS '
As in any data system, there is no actual limit to the sensitivity
, where there is a limit to is the acceptable % error rate
Opera data decode s/n can be lowered , but then comes the % false
decode increasing , you could counter argue that false decodes are
purely random , so are of no consequence .. some may
The design criteria for a correlation system remains the same ,
but where as in the data decode , logic decisions are called on to
extract the data from the coding , as in the case of Opera a
minimum of 40% random capture .
A correlated detection 'Operas Dynamic' , is a pattern match against
time , sharing the 40% minimum level , the information is presented
as a fixed pattern and the system desires a correlated match ,
much the same as the bombe and colossus at station X , the system
also seeks to minimise the retained call list , the 8 day period
was set to compliment the band activity ,
The design challenge is to produce a match with enhanced
sensitivity, whilst maintaining a 'credible' false detection level
compared to a data decode. In the case of Opera dynamic , this is
set at -5 bB with a commensurate low level of false detections ,
Zero false detections from a group of 14 stations in a 4 hour period ,
subjected to high band noise conditions , is a indication of the
level of the system design
As correctly noted random events do conspire to produce false
detections , the confidence level for a stand alone non web-linked
RX is indicated by the recovered s/n level, the user should refer to
tx time start etc
However, being a dynamic system, once more than a single Rx is
linked via the web, then cross correlation of the detection times ,
enables a very robust statement to be made , as to the authenticity
of the result ... noting unfortunately, the psk-map shows all
detections , where as the web-linked local copy is provided with the
system evaluation ''??'' for a false or only 'single' detection
The evaluation of the Dynamic detection , is based on the system
parameters and detection algorithms performance, providing a modern
robust fully automated addition to the established Opera system . The
only auto configuring dual mode low s/n system available to the MF
and LF user group.
73-G,
--------------------------------------------------
From: "Alan Melia" <[email protected]>
Sent: Thursday, May 21, 2015 5:47 PM
To: <[email protected]>
Subject: Re: LF: MF 630m: False Decode or Real?
A couple of comments on the statistics......this has little to do with the
decode process itself.
The timing only goes to show that the "decoded" call sign was active at
the time NOT that he was really received, but might have been.
Graham thinks that because there were no more false decodes in the
previous 24 hours this one MUST be good. Statistics dont work like that !
If it is a million to one for a false decode you must expect at least near
to a million decodes before another false one. On the contrary 2
questionable decodes of the same call would be extremely unlikely to be a
random correlation so would provide good evidence the call was actually
received and decoded properly. The probability of one true decode but not
a second due to fading is very (very) low but not impossible.
Next though many do not seem to appreciate it, statistics dont "prove"
anything, they provide a measure of the probability of an expected
outcome. You determine both the probability of the expected outcome and
also the probability it could be just random chance. The bigger the
difference then the more likely the expected event occured. This is what
makes a second decode so powerful.
People DO win the lottery every week! So false decodes do happen by
chance.....but the same lottery winners rarely win two weeks in
succession.
Alan
G3NYK
----- Original Message -----
From: "Mike Dennison" <[email protected]>
To: <[email protected]>
Sent: Thursday, May 21, 2015 5:04 PM
Subject: Re: LF: MF 630m: False Decode or Real?
Very likely a false decode. The clue is the two question marks. In my
experience, almost every decode with question marks is false, and
there are very few cases where these particular decodes are useful. I
believe it would be helpful not to display these at all.
Having said that, the deep search facility in Opera is very useful
and has revealed some very interesting propagation information. But
only using the decodes without the question marks which, again in my
experience, are almost always genuine.
Mike, G3XDV
===========
Just popped up at my RX:
17:52 477 VK3ELV de DF1VB/3 Op8 Deep Search ?? 16348 km -37 dB in
Dortmund with 140w + Top loaded L 18m vert 80m horz
Any comments welcome
73, Jochen
--
-= DF1VB =-
-= KH2MM =-
Jochen Althoff
+49 171 2020206
"The wireless telegraph is not difficult to understand.
The ordinary telegraph is like a very long cat.
You pull the tail in New York, and it meows in Los Angeles.
The wireless is the same, only without the cat."
(Albert Einstein)
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2015.0.5941 / Virus Database: 4347/9830 - Release Date:
05/21/15
|
