Re: LF: Octo-soundcard for the Raspi, another question

[Markus Vester <markusvester@aol.com>]

Hi Stefan, Paul,

this seems to be a weak effect: a sub-percent amplitude variation of a -50 dB FS carrier would correspond to less than an LSB on an absolute 16-bit scale. It's surely too small to compromise sensitivity of a real-world VLF receiver.

First I thought it might be flicker noise on the ADC reference voltage but why should it affect only three of six channels? Now I am wondering if it might be an EMC issue, with undesired feedback from the digital data transfer to the analog inputs. With USB soundcards, the data packets are usually sent with 1 ms repeat rate, and we have sometimes seen an artifact consisting of weak 1 kHz spaced spectral lines. I do not know about the transfer protocol of the Octo soundcard which is connected to the RasPi interface pins. In case the packets are organized similarly, we might experience the same problem, so 1 kHz or 0.5 kHz might have been just an unlucky choice for the test signal frequency. Perhaps you could try again at some odd frequency? Also a test at higher amplitude (e.g. -20 dB FS) would tell us whether it is a multiplicative (ADC scaling) or additive (EMC) error..

BTW This week I have received my Octo soundcard from the U.S. But I guess it will take some time until I have learned to install it on the RasPi and run vlf-tools under Linux. The warm summer is not a good time to bury oneself in the shack ;-)

Best 73,
Markus (DF6NM)

-----Ursprüngliche Mitteilung-----
Von: Paul Nicholson <vlf0403@abelian.org>
An: rsgb_lf_group <rsgb_lf_group@blacksheep.org>
Verschickt: Sa, 8. Jul 2017 19:40
Betreff: Re: LF: Octo-soundcard for the Raspi, another question

Stefan wrote:

> http://www.iup.uni-heidelberg.de/schaefer_vlf/VLF/record.tar

Interesting. The first thing I notice is some curious effects
on the background noise, on ch1 only, eg

vtcat -S49 -E2 170708-095003 | vtrsgram -ox -g200 -- -:1

(ctrl-C in the shell to quit)

or with vtsgram which uses sox as a back-end for the
spectrogram

vtcat -S49 -E2 170708-095003 | vtsgram -p200 | display

There are lots of these: For a very wide image

vtsgram 170708-095003 -p50 | display

(Use the pan icon for panning and q to quit).

Looking closely at the one at 24 seconds

vtcat -S23.5 -E1 170708-095003 | vtrsgram -ox -g200 -- -:1

and

vtcat -S23.5 -E1 170708-095003 | vtscope

gives http://abelian.org/vlf/tmp/170708a.gif

We can just see a slight change to the ch1 amplitude
which lasts about 100mS. Can't really see anything
obvious on the scope when zooming in on the start,

http://abelian.org/vlf/tmp/170708b.gif

The amplitude just seems to go up a tad at about 442 mS.

Lets look at the phase during that same 1 second window

vtcat -S23.5 -E1 170708-095003 |
vtnspec -f1000 -w0 -r20 -k |
awk '{printf( "%s %.1f\n", $1, atan2($4,$3)*180/3.14159)}'

produces

1499507427.027 121.1
1499507427.077 120.9
1499507427.127 120.8
1499507427.177 120.6
1499507427.227 120.5
1499507427.277 120.3
1499507427.327 120.2
1499507427.377 120.0
1499507427.427 119.9
1499507427.477 119.7
1499507427.527 119.6
1499507427.577 119.4
1499507427.627 119.3
1499507427.677 119.1
1499507427.727 119.0
1499507427.777 118.8
1499507427.827 118.7
1499507427.877 118.5
1499507427.927 118.3
1499507427.977 118.2

Just shows a steady phase drift, we're not quite on frequency here
with this recording - did it go through vttime? I see from

vtstat -i 170708-095003

that we have

sample rate: 24000, correction 0.99997722

The correction should always be 1.00000000 if the stream has been
through vttime.

Curious. Signal goes up, background noise drops. I wonder if
there is a dodgy input capacitor or dry joint. Not sure about
that explanation, it would be odd to see the same on more than
one channel.

Wonder if the effect occurs at the same time on ch1 and say ch4?
Or instead the effect is independent on the effected channels.

I will set my Octo up with the same test and look for the same
effect here.

--
Paul Nicholson
--