Hi Markus, Group,
Congrats again on this work. I was very satisfying to be part of it.
I'll follow though on your advice. The SR is now calibrated at 12k.
Another option is raspi. It is already set to ntp and it does not
use a silly operating system.
At the moment I must QSY to the uni to pick up a friend. We are going to
Signal Hill to fly a kite in honour of Marconi!
73
Joe
On Mon, 12 Dec 2016, Markus Vester wrote:
Hi Joe, LF,
after obtaining that first decode, I was able to improve the result
significantly by tweaking time and frequency offsets. However during this
optimization, the little blue peak in the spectrum-of-squares which had
initially guided me disappeared completely. So this peak must have been just a
lucky coincidence. Starting from these parameters, I could then get decodes
from the 4 and 5 UT transmissions as well:
file df dt EbN0
hhmm Hz s sym carr
0408 -9.8832 556.1 2.6 -0.6
0508 -9.8832 556.0 -0.7 -3.5
0608 -9.8831 555.9 5.4 0.5
So we know now that the PC clock was around 3.4 s late at the time, and your
DDS pilot is 0.8 mHz below 137.5 kHz (albeit this assumes 12000.0 Hz samplerate
which may not be correct).
Let me suggest the next steps on the receiver side:
1. Align the PC clock to UT within say 0.2 s. There are several options for
NTP: Here I am using a small utility called SNTP which G3PLX gave me some years
ago, which avoids clock jumps by gradually adjusting clock speed. Wolf's rsNTP
also worked fine for me, and is able to include a deliberate offset to
compensate for long sound path latency. I haven't tried Dimension-4 software
but it has been employed successfully by a number of LF amateurs. Last not
least IZ7SLZ made his own clock-setting utility using serial NMEA from a GPS.
2. Use the calibrated 12 kHz samplerate for decoding. This will minimize
residual symbol-timing errors, and allow us to accurately predict the frequency
offset for other transmit frequencies.
3. Let SpecLab's frequency offset detector track the pilot. This will make the
decoding process more comfortable, because we won't need to measure the RX
frequency offset individually for each run.
All the best, and thanks again
73, Markus (DF6NM)
-----Ursprüngliche Mitteilung-----
Von: Markus Vester <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: So, 11 Dez 2016 11:21 pm
Betreff: LF: VO1NA received my EbNaut
Hi Joe,
we made it! But only with a lot of luck.
Using default frequency and samplerate parameters, I first looked for the DDS
pilot in your nine files and noted the frequency offsets:
file pilot
hhmm Hz
2208 .1157 (pilot turned on half-way)
2308 .1148
0008 .1147
0108 .1149
0208 .1156
0308 .1160
0408 .1162
0508 .1161
0608 .1161
Thus I concluded that your rx might have small enough drift over 22 minutes.
Then I took the file for my last transmission (12110608.txt) and calculated
nominal frequency and start time offsets:
freq: +0.1161 - 10.000 = -9.9839 Hz,
time: 6:00:00.3 - 5:50:40.92 = 559.38 s.
You said that your clock was about 5 seconds slow, so I started searching around 554.4 s
offset. The signal is too weak to be seen directly, and I hadn't sent any plain carriers
before or after which would have helped finding at least the right frequency. After some
playing around I focused on one of the nearby small blue peaks, and tweaked both offsets
to center and maximize it more, but no decode happened. Then I tried shifting the timing
in 2 second whole-symbol steps. Each time I waited for the decoder to work it's way
through all the phase combinations, loosing more and hope with only gibberish decodes.
Finally near the end of the third long run, suddenly my message "HELLO LF"
stared back at me: Oh my god, we've really made it!
Congratulations, Joe!
Let's call it a day now,
73, Markus (DF6NM)
Re: LF: EbNaut tests 137.490 kHz/OFFLINE
Von: Markus Vester <[email protected]>
An: jcraig <[email protected]>
Datum: So, 11 Dez 2016 8:53 pm
Hi Joe,
thanks for the files! I have successfully downloaded them from your site, and
will now start looking for possible signals.
Best 73,
Markus (DF6NM)
-----Ursprüngliche Mitteilung-----
Von: jcraig <[email protected]>
An: Markus Vester <[email protected]>
Verschickt: So, 11 Dez 2016 6:50 pm
Betreff: Re: LF: EbNaut tests 137.490 kHz/OFFLINE
Hi Markus,
The files are uploaded to http://www.ucs.mun.ca/~jcraig/ebnaut/
Lets hope my inexperience hasn't messed up this experiment completely!
The screen dump appears to have been made 1317 utc this morning and
the computer is set to utc but is slow by ~5s. These brief carriers
are eigmatic -- I don't know what causes them.
The pilot was generated using an ad9851 without the x6 multiplier
and a 10 MHz clock. It seems the raspi python maths retains
the full 32 bit precision for the frequency word. A 0.5 was
added before the ratio was written to an integer from a float to
minimise the rounding error.
Please let me know if you need anything else. I'll review the
hints you sent. Thank-you!
73 & Good Luck!
Joe
...
|