Hi Markus,

Convinced that the oscillator had drifted since calibration 2 years ago,

Thanks for the tips on getting started on EbNaut.  I'll tinker with it
and hopefully make some progress. Our cat seems anxious lately!

73
Joe

On Thu, 12 Nov 2015, Markus Vester wrote:

Hi Joe,
 
an interesting change happened last night: as far as I can see, your frequency 
has shifted to nominal 137477.000 at 22:45 UT, and appeared to be rather stable 
thereafter.
 
Regarding EbNaut transmit: I haven't tried it on our RasPi yet. Both Domenico 
and I have run the Windows version of ebnaut-tx, in conjunction with 
USB-to-serial converters. RS-232 lines are normally current-limited to 20 mA so 
they can be directly connected to the DC-coupled IF input of any ring mixer. On 
137 kHz, I used a Minicircuits ZAY3. As I have a linear PA I inserted a simple 
500uF/100ohm shaping filter to reduce keying sidebands, but this may not be 
essential. This mixer won't work on VLF 8.3 kHz so I used an old telephone 
relay there, but due to contact bouncing the transitions were less than 
perfect. Domenico uses an SSB exciter on LF, and he made a dedicated diode 
mixer with audio transformers to apply PSK on 1502 Hz.
 
Our current receiving method re-uses opds32 data files and is thus limited to 
sequence durations of a bit over half an hour. For example, using EbNaut 8K19A 
encoding with 1 second symbols would fit 31 characters into 29.3 minutes, and 
the message could be repeated half-hourly. Yesterday your opds carrier was 
weakly visible until midday in Europe, so daytime TA decodes might be possible 
using slower rates (e.g. 5 seconds).
 
Our cat seems to be very curious about radio, so I thought he might feel a 
desire to communicate with his fellows overseas. Rumour has it that at least in 
one instance, Uwe's cat successfully managed to take control of the DJ8WX VLF 
transmitter ;-)
 
All the best,
Markus (DF6NM)




-----Ursprüngliche Mitteilung-----
Von: Markus Vester <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: Mi, 11 Nov 2015 6:12 pm
Betreff: Re: LF: VO1NA frequency

  Joe,
  
 more of the same from last night, this time with frequency and level plots. 
Again we see some frequency variation of similar size as yesterday. It is not 
clear how much of that is due to propagation, I only think that 
the rise near the end is due to Doppler from the descent of the ionosphere 
at sunrise. Russian Loran lines on 137475 and 137482.5 seemed to have less 
overnight variation (~ 0.1 mHz), but then of course the intra-European path 
from Slonim is very different from yours. 
  
 Regarding EbNaut carrier analysis: for some reason (lack of coffee?), I only 
saw Domenico's earlier mail after I had sent mine, so please excuse me for 
basically having described the same procedure again. Anyway it's good that 
we both arrived at the same conclusion: If Joe started sending EbNaut PSK 
with his oscillator as it is now, we would very likely be able to decode it 
here. 
  
 Joe so all you need now is a diode ring-mixer for PSK modulation and Paul's 
ebnaut-tx program. Symbol timing will be derived from the PC clock, so 
with NTP available there's no need to involve GPS at all. If Paul brought his 
LF receiver back to life, he could probably copy you even if you reduced TX 
power to 0.1 Deccas.
  
 How many cats do you have? We want to know names, colours, sex and age for 
all of them ;-)
  
 All the best,
 Markus (DF6NM)
  
  
 PS. SpecLab question: Wolf is there a way to extract peak frequencies from 
past spectrogram lines (semi-)automatically, without having to 
manually move and read the mouse cursor over each point, e.g. something like 
a peak_f(f1,f2, time) interpreter function?

  
-----Ursprüngliche Mitteilung-----
Von: Markus Vester <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: Mi, 11 Nov 2015 1:24 am
Betreff: LF: VO1NA frequency
          Hi Joe,
   
  the attached plot shows the  frequency stability of your opds trace received last  
night http://df6nm.bplaced.net/LF/opds32_151110_0736.png .  Each data point is the interpolated peak of a  
single 438 uHz FFT (i.e. 38 minute window). The average frequency seems to be  0.7 mHz below nominal, and 
there are small temporal variations up  to 0.5 mHz in either direction. Though some of that is  
probably attributable to noise and propagation, I believe that there  is also a bit of inherent instability 
in the frequency  source. Actually a few parts per billion is quite good for a free  running OCXO 
(assuming the "D" stands for  "double oven" and not "discipled"), and that the 
calibration was done over a 10  MHz HF link.
   
  Is this good enough for EbNaut? During the  experiments with IZ7SLZ we found 
that for full sensitivity the phase variation  over the EbNaut sequence 
duration should not be much more than 90°. Thus  for a half-hour message the 
frequency should be ideally constant and known to  about 0.14 mHz. If there is 
more variation we often can still  get decodes, but we will have to find 
the best frequency offset by trial  and error, which can be cumbersome if the 
signal is weak.
   
  I also fed opds data files from your Sat/Sun  transmissions to the EbNaut decoder. The 
idea was to see if your Opera  transmissions would actually produce the 
all-asterisk decode which is  expected for a non-PSK-keyed straight carrier, and to 
observe the symbol phase  evolution over the 38 minute duration. Starting midnight UT, 
one decode was attempted every 30 minutes,  using a setting for 464 raw symbols (4 
characters) and 5 seconds per  symbol. A frequency offset of -0.6 mHz was selected to 
obtain  flat phase during the first decodes, and that offset was kept for  all later 
attempts. Despite the Opera gaps  which degrade or drop half of the available 
symbols, 12 out of 13  decode attempts produced the  correct "****" 
message, e.g. http://df6nm.bplaced.net/LF/r0317.png  . However at times the frequency 
drift was quite visible on the phase  plot, e.g. in 
http://df6nm.bplaced.net/LF/s0317.png (flat  phase around 3:17) versus 
http://df6nm.bplaced.net/LF/s0347.png (upward  phase slope around 3:47). 
   
  Scaling down to VLF, deviations from the same  oscillator will be 16x smaller 
(~30 uHz at 8.3 kHz). This will be probably good  enough even for narrow FFT 
bins, corresponding to several hours coherent  carrier integration or slow 
EbNaut decodes.
   
  All the best,
  Markus (DF6NM)


    From: [email protected]
  Sent: Saturday, November 07, 2015 3:45 PM
  To: [email protected]
  Subject: Re: LF: VO1NA opds-32




LF Group,

Again, thanks to SV8RV DF2JP PA7EY and DF6NM for  the reports.

In my last post I reported the drift was constrained to  within 1 Hz.
This should be 1 mHz.  Thanks Markus for your assurances  that this
should be adequate for EbNaut which I hope to use to send send our
cat's name on LF.  The little curls appear often on the QRSS traces  on
your high res spectra.  Doppler variations were removed from the 10  MHz
WWV signals with harmonic regression when the DOCXO was calibrated
so  this effect is seen on LF but not as strongly.

After  reading about  EbNaut on Paul Nicholson's pages it seems I have
much catching up  ahead.  Fasinating stuff!

Can the EbNaut phase inversions be done  without an absolute time
reference?  I'm cheap and lazy so hoping GPS  will not be needed.

The Opera will be doing an encore tonight and if the  wind stays low
a tower climbing expedition is planned for the  afternoon.

73
Joe VO1NA
...