Jacek wrote:

> what noiseblanker settings did you find to be the
> best on 2970Hz?

I've been trying a range of settings.

I reverted back to simple (daytime) options

  vtfilter -ath=5,bw=1 -h bp,f=3370,w=1000 |
  vtblank -a25 -d0 -t100 -v

(using vtblank from the latest version v0.8 of vlfrx-tools)

for the current run of stacking trials.  Blanking factor is
less than 5%.  There is not so much sferic energy in this
band and I'm only gaining about 5dB or so.

I don't know the best settings, they may depend on how
effective the hum filtering is.    I did some trials with
injected signals but the optimum setting didn't work very
well on vlf6.  I really don't know much about sferic
blanking anymore - I used to do but now I'm reduced to
guesswork.

Stefan wrote:

> 3 sigma.

Nothing to get excited about, 3 sigma is nothing really.
In every spectrum I run there are 3 sigma peaks, but after
hundreds of 2970 spectra it's the first 3 sigma that's on
the right frequency and it also happens to have about the
right bearing (although the relative phase is funny which
counts against it).

> How much sigma do you need to make a secure claim

5 is good.  I'll take 4 if it has the right bearing
and relative phase.  Of course, it has to be exactly
the right frequency.

> Thanks for the overview of the signal received in
> Bielefeld. Quite strong and stable!

Yes.

I can't measure the phase unfortunately, too many timing
breaks, same problem with Cumiana.   About 65 breaks
per day on vlf6, about 40 per day on vlf15.   A long
standing problem which is no problem for natural radio
but prevents reliable long integrations.   Here, I have
only one timing break in several months and that was
on 1st Jan due to the leap second.  The one before that
was a power cut in a storm last autumn.

With this 2970 test and trials of convolutional codes my
poor workstation is getting hammered!

Still haven't produced a rate 1/16 K=19 code that's any
good but I've made a little progress in calculating code
performance, eg

 http://abelian.org/vlf/tmp/170321a.gif

It took 9 CPU hours to calculate the 8K19A curve but
over 200 CPU hours to accurately measure the performance,
so that's some gain although it's taken me 18 months to
work out the math. Wish I'd paid more attention during
math at school!   Unfortunately the calculation for
higher rank leads to some hideous combinatorics. I have
to find a better way.

One thing for sure with this VLF hobby, we're never short
of interesting challenges - everywhere you look there
is something new to try.

--
Paul Nicholson
--