[Lf] LORAN-C based frequency standard

[Andre' Kesteloot]

Stewart Nelson wrote:

> Hi all,
>
> Paul W1VLF wrote:
>
> > I  was wondering if anyone on the groups has experience with
> > LORAN-C receivers.  There is a LORAN-C station close to me...
> > I believe Nantucket Island.
> > What I would like to know if anyone has gone into A LORAN-C
> > receiver and found a point to sample the 100kHz that is not gone
> > thru a mixer yet.  I need to use the 100kHz as  a freq. reference
> > to lock my DSP station's receiver's oscillator.
>
> I have recently spent considerable time examining LORAN-C signals.  I
> concluded that it should be possible to build a very simple PIC-based
> frequency standard (total cost under $100 including oscillator), which
> would provide short term frequency accuracy < 1E-9, with long term
> timing within 100 ns of UTC.  Although this short term performance
> would be greatly inferior to (OCXO based) GPS standards, it is
> adequate for many purposes.  Being within 0.2 Hz on two meters is not
> all that bad.  A LowFER signal would stay within 7 degrees of the
> proper phase, for any length of transmission.
>
> Compared with the GPS unit, it would offer the following advantages:
> * Operate with small indoor loop antenna.
> * Two minute "warm up" (vs. about one hour for GPS / OCXO.)
> * No external receiver or oscillator to buy.
> * Uses about one watt total.
> * All home brew.
>
> It would still be possible to add an external OCXO and get performance
> comparable to GPS standards.
>
> John Davis wrote:
>
> > Alas, a "trivial" receiver is not going to be much use with LORAN,
> > either.  Even if one can depend on having a local groundwave signal,
> > that won't be  the only one present.  Nor will that local station be
> > transmitting pulse groups for only one chain!  One must still be
> > quite selective about the pulses used.
>
> I believe that the needed "selectivity" can be provided in software,
> so I built a very simple receiver and connected it to a PC to see how
> it would perform.  It was much better than expected.  The local
> station here (Fallon, NV, about 53 miles away) is admittedly not dual
> rated, but I believe that a simple median filter could be added to
> deal with that problem, and that the system should work well for
> anyone within about 300 miles of a LORAN-C station.
>
> First, my balanced loop was tuned to 100 kHz.  The bandwidth was too
> narrow and the preamp overloaded, so a resistor was added across the
> loop.  I think the Q is now about 40, and the LORAN pulses out of the
> preamp peak at about 4 volts p-p.
>
> Now, the main ingredients of most receivers are amplifiers and
> filters, but the sound card clips at about 0.5 V p-p, so only -18 dB
> of voltage "gain" was needed, and a couple of passive RC stages
> easily remove the 200 kHz and higher products from the mixers.
> The 10 MHz Rb oscillator is divided by 25 and drives a dual flip-
> flop connected to generate quadrature signals at 100 kHz.  These
> then drive transistor switches connected as DBMs.  You can see
> the receiver output for a typical master "A" pulse group at
> http://www.scgroup.com/lowfer/pulsgrup.gif .
>
> The I and Q receiver outputs feed stereo sound card inputs.  This
> experiment does not attempt to actually lock the Rb to LORAN.  It
> merely logs apparent time differences between LORAN and the local
> clock.  The resulting graphs are examined to evaluate how well e.g. a
> TCXO or OCXO might fare if locked to the LORAN signal.  The software
> samples each channel eight bits at 22.05 kHz.  It sums six consecutive
> values (containing one pulse), adjusting polarity for position in the
> group and A/B fields.  Eight pulses are accumulated per group, and 100
> groups are averaged (9.94 seconds at the local GRI).  The phase is
> computed from the I and Q averages, and the result is logged to a
> file.  Because the sound card has an inaccurate and unstable timebase,
> the software compares the energy in the first three samples of each
> pulse with the last three, and uses the difference to adjust the
> number of sample times to the next pulse group.
>
> Plots were made using averages of 600 pulse groups, approximately one
> minute.  At http://www.scgroup.com/lowfer/loranorig.gif you can see
> results from the first run.  The X axis is in hours, the Y axis in
> microseconds.  Zero is the initial phase; positive numbers represent
> LORAN advanced relative to LO.  It appears that the LO is way off
> frequency.  You can't see much else.  In fact, the frequency is only
> about 4.4 uHz high, but other errors are small by comparison.  What a
> difference from the WWVB data!  I tried adding a suitable slope to be
> able to see the other variations, but my imperfect mixers severely
> distorted the results.  So I used the receiver to carefully adjust the
> frequency pot on the Rb oscillator.  After a few tries and with some
> luck, I managed to get the LO within 30 nanohertz of the average
> received LORAN signal!  It soon became clear that variations in
> outside temperature were affecting the received phase, so I added
> temperature to the plot.  You can see the results with the adjusted
> oscillator at http://www.scgroup.com/lowfer/loranadj.gif .  The left
> hand scale is microseconds - the entire plot is only 200 ns high.  The
> right hand scale is degrees C, and the X axis shows hours relative to
> local midnight on March 2.
>
> I had failed to realize that, even with a Q of 40, tiny changes in
> loop tuning would cause large (for this purpose) phase shifts.
> Although I used 1% silver mica capacitors to tune the loop, a final
> adjustment was made by putting a 0.02 uF ceramic in series, and that
> caused most of the instability.  A quick calculation showed that if
> dimensional changes of the loop were the only temperature effect, the
> shift would have been about 25 ns, still significant.  A real design
> would reduce Q to about 10 (still enough to reject AM broadcast,
> etc.), use a temperature compensating tuning capacitor, and bring the
> loop indoors.  Too lazy to fix the system, I compensated temperature
> in software; http://www.scgroup.com/lowfer/lorancomp.gif has that
> plot.  The left hand scale is still microseconds, showing less than 60
> ns variation from all remaining causes.  The short term behavior is
> much better than GPS, although there are occasional sudden shifts of
> about 20 ns, which I cannot explain.  The red plot shows relative
> signal strength, using the right hand dB scale.  It is also
> temperature compensated; Lyle's preamp exhibits a gain coefficient of
> about -0.01 dB per degree C, actually quite good for an amp with no
> feedback.  Some of the sudden phase shifts are accompanied by
> amplitude changes, but others are not.  Any ideas about this are
> welcome.
>
> But even with these errors, a LORAN frequency standard using a short
> PLL time constant (so it would work with a relatively unstable
> non-ovenized crystal) would still perform better than a similar system
> using GPS, which would be degraded by SA.
>
> I am curious how many LowFERs might want a standard based on this
> technology, and would also like to hear from anyone willing to help
> with various aspects of the design.  All comments welcome.
>
> 73,
>
> Stewart Nelson  KK7KA
>
> To unsubscribe, send to MAJORDOMO at qth.net "unsubscribe lowfer" (Do not
> send to list!!) Send on list submissions to lowfer at qth.net