I should mention that for sferic blanking to be this effective
it is essential to band limit the VLF before the blanker.

A broadly optimum bandwidth is 3kHz centered on the signal
frequency.   Too wide and too much noise is modulated into the
signal band by the chopping action of the blanker.  Too narrow
and sferics are stretched which gives too much blanking.

It is also important to make sure that the 3kHz band passed into
the blanker doesn't contain any significant constant signals
such as mains harmonics.   Weak ones are OK so long as they
don't show above the 3kHz wide noise.  Therefore you may have
to precede the blanker with an automatic hum notch filter.

Getting the sferic removal working right is crucial to detecting
weak signals at VLF.   I'm sure that many reception attempts fail
because the blanking is nowhere near optimal.   It might be very
well worth doing some trials using a weak near field source with
amplitude set so that it is only just visible without blanking.
Record about 30 minutes of raw VLF with the test signal buried
in there, then use the recording as input for trials of various
settings with a bandwidth of 1/(30 mins).

In vlfrx-tools a typical detection pipeline is

 vtread -T2016-04-19_13:15,+2h /raw | # Select the signal (3 chans)
   vtmix -c 0.927,-0.375,0.5/-135 |   # Antenna mix and phasing
   vtfilter -h bp,f=8270,w=3000 |     # 3kHz filter
   vtblank -a1.2 -d0 -t1 |            # Sferic blanking
   vtnspec -f 8270.0025 -r 138.88e-6 -w0.05 > df6nm.160419a.dat

The blanker options are:

 -a 1.2     Auto threshold, 1.2 times the moving mean level
 -d 0       No dwell time
 -t 1       One second time constant for the moving mean

--
Paul Nicholson
--