Re: VLF: Trying the impossible, once more - symbol error rate for short data

[Markus Vester <markusvester@aol.com>]

Hi Paul,

when working with wav files that are too short, we find that Eb/N0 based on symbol error rate is usually shown too low. I think that you mentioned a while ago that although non-existing symbols are completely ignored by the decoding process itself, they are counted as "always wrong" in the error statistics. Wouldn't it be more fair to assume 50% error rate for missing data?

Best 73,
Markus


-----Ursprüngliche Mitteilung-----
Von: Paul Nicholson <vlf0403@abelian.org>
An: rsgb_lf_group <rsgb_lf_group@blacksheep.org>
Verschickt: Fr, 13. Okt 2017 0:38
Betreff: Re: VLF: Trying the impossible, once more

Stefan wrote:

> f = 6470.100000 Hz
> Start time: 12.October.2017 06:00:00 UTC
> Symbol period: 180 s
> Characters: 1
> CRC bits: 0
> Coding 8K19A

A great signal in Todmorden, Eb/N0 = 15.9, phase -114.3.

Has the potential to go a long way! It actually decodes
correctly here after only 980 seconds of signal, about 5.5
symbols out of 192.

--
Paul Nicholson
--