Re: LF: New opds version

[Markus Vester <markusvester@aol.com>]

Moving on from opds2h5c to opds2h6 a couple of months ago has indeed significantly reduced the number of false positive detections on my LF opera monitor, with no apparent loss of sensitivity: http://www.df6nm.de/opera/opds.htm . However with default settings, partial correlations to strong real Opera signals can still occasionally lead to false spots on an adjajent frequency. This can be fixed by adding a line to the opds.ini configuration file:
 minpeakspacing=200
allowing opds to associate any spurious peaks within 200 FFT bins (~ 0.09 Hz at LF) to the original signal.
 
Some opds users are still using Dropbox to upload their results. Due to a recent policy change regarding the Dropbox public folders, these will no longer be uploaded to the web, and any links like dl.dropboxusercontent.com/... etc. are now going into digital Nirwana. I have replaced Dropbox by FTP to my webspace. The FTP transfer can be automated by adding a line to cleanup.bat like

@ftp -s:ftp_opds.txt >ftp_log.txt

and preparing an ftp.txt command file containing something like
 
open <myftpserver>
<myaccount>
<mypassword>
cd opds
bin
put opds32.txt
put c:\spectrum\opds32.png
bye

Best 73,
Markus (DF6NM)


-----Ursprüngliche Mitteilung-----
Von: Markus Vester <markusvester@aol.com>
An: rsgb_lf_group <rsgb_lf_group@blacksheep.org>
Verschickt: Mi, 18. Jan 2017 15:02
Betreff: LF: New opds version

Dear opds users,

I have recently discovered a minor bug in opds2h5c which could sometimes produce systematic false detections from partial correlations after the end of a strong Opera signal. This has hopefully been fixed in version 2h6:

 http://df6nm.de/opera/opds2h6.zip

This version should be completely compatible with existing configurations and ini files, so all you have to do is start opds2h6 instead of opds2h5c.

Differences are:
- potential false partial correlations are cross-checked against detections on the same frequency during the last 32 minutes,
- to further reduce possible false detections triggered by strong non-Opera signals, the correlation threshold is now raised with signal strength. By default, it goes up linearly from 15 to 19 dB, starting at SNR -41 dBOp. If desired, this behavior can be modified by dedicated opds.ini entries, 
- the bug in the default filename pattern has been corrected, so the "pattern=*.txt" ini-line is no longer needed,
- unused parts of the code have been removed, resulting in a slightly smaller source and executable.

BTW For the old 16-bit QuickBasic versions before opds2h5b, I had recommended to prevent CPU-load peaks by running the process at lower priority. This seems to be no longer necessary.

Best 73,
Markus (DF6NM)