Hi Markus and all,
I'm quite interested in this idea; perhaps we can work together
to develop the code. We might call it the LAMP (Loran Analysis
for Monitoring Propagation), but other suggestions are welcome.
I haven't written any real code yet, but processed some .wav files
with a simple perl script and plotted the results with Excel.
You can see a sample at http://www.scgroup.com/ham/h9610.gif .
I used a one-minute recording made by Dexter W4DEX in Stanfield,
North Carolina. The horizontal scale is samples from the beginning
of the recording, modulo the GRI "tuned", 9610 in this case. The
vertical scale is arbitrary voltage units. The four colors show
correlation with master and secondary codes, with both A/B sequences.
Layout of the 9610 chain can be found at
http://www.megapulse.com/pix/chain/9610.gif . The magenta peak at
1592 is the master in Boise City, Oklahoma. The big yellow peaks
at 307 and 641 are Raymondville, TX and Grangeville, LA, respectively.
Cyan peaks at 2069 and 2618 are Gillette, WY, and Searchlight, NV.
This last one is 3115 km from W4DEX. You can't see Las Cruces
because of a bug related to wraparound.
My best DX is Ejde, 5760 km from W4DEX. It was just slightly above
the highest noise peak for the 9007 chain. Unfortunately, I couldn't
positively identify stations such as Sylt or Lessay.
To make this system really useful, we need IMO about 10 dB more S/N.
I have some ideas on how to get it, and I'd very much like to hear
yours. For starters, we can integrate for more than one minute.
Unfortunately, this must be partly noncoherent, because I think
that the skywave path is generally not phase stable for longer than
a couple of minutes. We could use a filter properly matched to
the receiver's impulse response. My script has only a crude two
point equalizer. For dual-rated stations, we could coherently
combine pulses from both GRI's (see below). And, most complex,
we could use an adaptive equalizer to cancel most of the local
QRM. W4DEX is only 261 km from the Carolina Beach station;
the present script simply goes deaf during those pulses. The
perl code is at http://www.scgroup.com/ham/loran1f.txt .
You can see that it is pretty crude - I used Cool Edit to convert
the .wav to 32000 Hz samples and to text, and determined the true
sample rate and BFO frequency manually. Sorry that there are
no comments.
Not only do LORAN transmitters use atomic clocks, but they are
monitored and kept within 200 nanoseconds of UTC. The long term
frequency error is zero! Also, every LORAN master conceptually
emitted a pulse on January 1, 1958 at 00:00:00 UTC. So if you
know the time within 1 GRI, you can tell the exact time. Better,
by looking at the timing from a dual-rated station, you only need
know the time within a few seconds. With two dual-rated stations,
within several hours, and with three, within about a year. So,
by carefully examining the recording, I determined that the first
recorded master pulse from the 9960 chain was the one emitted from
Seneca on March 12, 2001 at 01:28:31.0544 UTC. I then used this
information to predict exactly where pulses should appear from
stations in various other chains. If within ground wave range,
the match was always within 20 microseconds, so any additional
sky wave delay could be judged quite accurately. Also, the
amplitude of the signal from the closest station is probably
consistent within 0.1 dB, so the system could continuously
self-calibrate and give accurate field strength readings, too.
How should we proceed from here?
73,
Stewart KK7KA
----- Original Message -----
From: <[email protected]>
To: <[email protected]>
Sent: Tuesday, March 20, 2001 4:38 PM
Subject: LF: Loran DX
Hi Wolf and all,
triggered by John VE1ZJ's recent remarks on Loran as a skywave propagation
monitor, a couple of weeks ago I took a deeper look at what could be
received. John's "www.G4CNN.f2s.com/Loran_lines.htm" pointed me to the list
at "www.megapulse.com/table.html". With this at hand I tried to identify the
lines I could see on Argo around 100.0 kHz.
The key to their frequencies is the "GRI" (group repetition interval), which
is the number of 10us carrier periods between two repetitions of the
modulating pulse groups. Each group consists of 9 or 8 pulses, 1 ms apart.
Some of these pulses have an alternating phase, so that the periodicity of
the pattern is actually two times the GRI:
Master Secondary
++--+-+- + +++++--+
+--+++++ - +-+-++--
Thus the frequency spacing of the lines is Df = 100kHz/(2*GRI), eg.
100kHz/(2*7499) = 6.6676 Hz for the Sylt chain. The Loran-C carrier frequency
is generated by atomic clocks and claimed to be accurate on the order of
10^-13.
The chains I could clearly observe here were
GRI Df Chain (Wolf's AM line)
/10us /Hz
5930 8.4317 Canadian East Coast
6731 7.4283 Lessay
7001 7.1418 Bo (140*Df = 999.8572 Hz)
7030 7.1124 Saudi Arabia S
7270 6.8776 Newfoundland East Coast
7499 6.6676 Sylt (150*Df = 1000.1334 Hz)
8000 6.2500 Western Russia (160*Df = 1000.000 Hz)
8830 5.6625 Saudi Arabia N
9007 5.5512 Eide
Then there were additional weak lines which were too close to 100 kHz, at
offsets of 1.52, 3.04 and 4.56 Hz. Their explanation is a little more subtle:
Many loran stations are "dual-rated", they transmit in two chains with
different GRI's. In case of a collision between two pulses that would have to
be sent simultaneously, one of the pulses is simply left out. These dropped
pulses occur at the beat frequency between the two GRI's. For the Sylt
station, these "intermodulation" lines are multiples of
(100kHz/6731-100kHz/7499) = 1.5215Hz.
There were even more lines I could not identify, eg. on 5.14, 5.90, 7.24,
8.18, 8.95 Hz. This made me wonder if the table is really complete, as stated
by megapulse. Also, I can't explain Wolf's observed 999.96 Hz.
The fun got even more interesting when I went to time domain. Using a
programmable divider clocked by 100 kHz, I generated 2*GRI trigger signals
for a digital oscilloscope in 128-averaging mode, and viewed the SSB output
tuned to 100.0 kHz zero beat (thus allowing phase-sensitive averaging). With
this setup I could see the individual pulse groups of distant chains grow out
of the noise, identify their stations and measure the time-differences. These
matched calculated great-circle distances with a surprising accuracy of less
than 100us, and it even worked fine for all five of the Saudi-Arabian
stations, up to 4642 km from here.
Thus I think that Loran-C can be used not only as a precise frequency and
time standard, but also as a powerful instrument for worldwide LF propagation
monitoring, aided by its ability to resolve propagation delays.
73s and happy experimenting
de Markus, DF6NM
|
