To: rsgb_lf_group@blacksheep.org
References: <16190d3155b-c8f-1c19b@webjas-vab095.srv.aolmail.net>
From: Paul Nicholson <vlf0403@abelian.org>
Message-ID: <319348e4-b241-c703-85c9-ba8598075f5c@abelian.org>
Date: Tue, 13 Feb 2018 21:03:10 +0000
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.3.0
MIME-Version: 1.0
In-Reply-To: <16190d3155b-c8f-1c19b@webjas-vab095.srv.aolmail.net>
Content-Language: en-US
Subject: Re: ULF: On the way to results... | vlfrx tools - advice needed...
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org


I have only time for a quick reply for now...

The timing and phase are not upset by passing the signal
through vtfilter, or vtresample, or any of the other utilities.
Of course if you give vtfilter an eqmap that specifies a phase
shift, then the phase will shift (but not the timing).
Ditto for vtmix if you give it a phase factor or complex
coefficient.

The filters produced by vtfilter -h options change only
the amplitude, not the phase!  They are not normal filters
in that respect.

You were lucky to get a good E + H sum without scaling amplitudes
or phase shifting!

I would expect that best results should be obtained by mixing the
antennas first, then apply the blanking (for the reasons you
mention).

But, I usually find that blanking the channels first and then
combining them, produces a slightly higher S/N, sometimes a couple
of dB higher.  I don't know why this is.

 > How can i mix E and H again from separate files?
 > Is it done with vtjoin?

Yes, but don't use lock-free buffers.  They're only good for 'slow'
signals, ie at roughly real time rates.   If you put a signal
through a buffer during post-processing, the supplier of the signal
might easily run faster than the consumer process and the buffer
will overrun.   You should never need a lock-free buffer during
post processing.   You can use the 'tee' program to split a signal
several ways, stuffing each copy into a pipe (named or not).
Look up 'Process Substitution' in the man page for bash, eg

  tee >(pipeline one) >(pipeline two) < input.vt | pipeline three

where the three pipelines are arbitrary lists of shell commands.

So assuming you've extracted E and H to signal-e.vt and signal-h.vt,
each is a single channel, and they both cover about the same time
range at the same sample rate, then

  vtjoin signal-e.vt signal-h.vt combined.vt

will give you a 2-channel combined.vt with E on chan 1 and H on
chan 2.  Then mix with

  vtmix -c coef-e,coef-h combined.vt > mixed.vt

where the coefs are your (possibly complex) mixing coefficients.

Do it in one step with

  vtjoin signal-e.vt signal-h.vt - | vtmix -c coef-e,coef-h | ...

where the dash tell vtjoin to output to standard output.

My usual post processing is in this order:

  vtread -T  (extract signal from recording)
  vtcat -p   (to pad over any timing breaks)
  vtfilter -e eqmap -h ... (antenna/rx phase and amplitude EQ,
                            and 3kHz bandpass before blanker)
  vtblank     (-a12 night, -a24 day)
  vtmix       (mix to form the antenna response, usually linear
               polarised cardioid but sometimes circular)
  vtmult      (mix down to baseband)
  vtresample  (often in two stages, down to 1Hz or 10Hz or so)
  ebnaut -d   (decode)

vtblank, by the way, blanks all signals together.  In other
words, if one channel exceeds the threshold, the blanking is
applied to all the channels.  This 'joint blanking' policy can
be changed with -e and -b options.  Joint is probably best.
If one channel is blanked but not another, then your mix is
briefly wrong during the blanked periods.

--
Paul Nicholson
--