> What do you consider suitable?

In order to squeeze out the maximum S/N, it looks like a
loop antenna is desirable in order to null as much
background noise as possible.

It must 'see' down to the natural VLF background, so a
basic sensitivity of say better than 5fT in 1Hz is
required.    Then the floor should be dominated by the
natural background.   This doesn't seem too demanding at
9kHz and it looks like people are having some success with
resonant antennas connected straight into the soundcard!

A pre-amp would enable a smaller loop to be used which
increases options for locating the antenna in a quiet spot.
Transformer isolation may be necessary to remove ground-
looped interference - even if the loop is balanced.

Sampling needs to be stable to better than 0.01Hz, which
may be a problem for low grade soundcards, especially if
timing is based on ceramic resonators rather than xtal.

Spectrum Lab contains tools for doing the initial conversion
rate calibration and also for making continuous correction
during reception by using a reference signal.  It is
important that the signal stays more or less within one bin
during the test.

It is worth also discarding the 5% or so of signal that
contains the strongest sferics.    May be worth recording
the test raw and post-processing to try various settings
for blanking and frequency binning.

If the signal crosses the Atlantic, it may be right on the
edge of detectability.   We used a pessimistic model for
the previous test and the signal was larger than expected.
But now we are probably using an optimistic model. Therefore
luck cannot be invoked and it is down to skill, ingenuity,
and probably patience too - a long integration will be
needed!

--
Paul Nicholson
--