[Lf] LF path stability; frequency references]

[Andre' Kesteloot]

Stewart Nelson wrote:

> This is an attempt to answer some of the recent questions about
> frequency standards, especially those locked to LF sources such as
> WWVB or LORAN-C.  I plan to make a Web page with this information, but
> would first like to get criticism/suggestions from the amateur
> community.
>
> It is obvious that weak signal low bandwidth digital communications
> can take place at much lower signal to noise ratios if accurate and
> stable frequencies are employed.  The basic questions are how much
> accuracy is needed for various ham purposes, and how can it best be
> obtained.
>
> One guideline is to make the station stability several times better
> than that imposed by the propagation path.  For example, a Lowfer
> single-hop nighttime skywave signal might exhibit a phase change of 30
> degrees over an interval of 20 minutes, under good but not exceptional
> conditions.  A 30 degree shift is the point where BPSK demodulation
> starts to seriously degrade.  If the Tx and Rx oscillators are each
> ten times better than this, then the bandwidth needed would be
> increased by at most 20 percent.  To avoid a requirement for the
> receiver to search for the signal, the frequency accuracy, as well as
> its stability, should meet this standard.  Three degrees in 20 minutes
> is about seven microhertz; at 180 kHz that's about 4E-11!  Of course,
> most of us would use a character length shorter than 20 minutes and
> tolerate a larger degradation; we might decide that 1E-9 is adequate.
>
> Libration fading over the EME path typically imposes a spreading of
> about 0.4 Hz on a two meter signal.  If we want 0.1 Hz accuracy (with
> careful compensation for Doppler), we need 7E-10, in the same ballpark
> as above.  Surprisingly, the accuracy requirement actually decreases
> somewhat on higher bands, because the moon appears rougher for shorter
> waves, and reflection from the larger area causes libration to
> increase faster than frequency.
>
> Lasercomm is much more forgiving, because an audio frequency
> subcarrier is used.  If we tolerate a ten degree shift of an 800 Hz
> tone over 20 minutes, that's about 3E-8.
>
> Now that we know how accurate we would like our signal to be, how
> can we get it?  One can either use a highly accurate oscillator, or
> synchronize a lesser oscillator to some broadcast signal.
>
> At http://www.rakon.co.nz/generated/232-241/233.html there is a
> good comparison of various oscillators.
>
> A cesium oscillator is overkill for any ham application I can think
> of.  They are very expensive; careful shopping for a used unit might
> yield something functional for about $2000.  They are also large and
> heavy, require lots of power and TLC, and have inherent wearout
> mechanisms.  But, if you have a profession or another hobby which
> demands exceedingly accurate timing, you can find a great guide to
> these units at http://phk.freebsd.dk/cesium/ .
>
> Rubidium oscillators are not cheap, either.  But demand for these has
> decreased recently, because better OCXO units have taken the low end,
> and GPS-synchronized standards the high end.  So there are lots of
> used units for sale on the Net.  For about $500, you can get one with
> an NIST-traceable calibration certificate and a meaningful warranty.
> For example, see http://used-line.com/Osc1101.htm .  I bought one from
> http://www.vigilante-electronics.com/ for $350.  It was only
> guaranteed to work when received, but I am very pleased, as my unit
> seems to only gain about four microseconds per day (much better than
> spec).  Your mileage may vary.  It also came with a matching connector
> and a great manual.  Such units are also often sold on Ebay.  Rb
> oscillators wear out; my unit might have two years (power on time)
> left on its lamp.  They need 10 watts or more, making them unsuitable
> for some portable applications.
>
> The newer high-end OCXOs are amazingly good.  If you can get hold of
> one, go for it.  They take less than a watt, and should last almost
> forever.  If you ever decide you need more accuracy, you can usually
> find something to synchronize it to.  Which brings me to the main
> point of this posting.
>
> You can synchronize an oscillator to various sources.  From a
> performance viewpoint, GPS is clearly the best.  For a few hundred
> dollars, you can have a clock which is always within 100 nanoseconds
> of UTC, with short term accuracy of 1E-11 or better.  Brooks Shera's
> excellent design can be found at
> http://www.rt66.com/~shera/index_fs.htm .  I can only think of a
> couple of reasons why you might want something else.  For GPS, you
> will probably need an outdoor antenna.  Although the antenna is tiny,
> it still means running a cable, and worrying about weatherproofing,
> lightning protection, etc.  Also, it would be very difficult for an
> amateur to build a GPS receiver from scratch; for a fully homebrew
> system, LF is the way to go.
>
> The obvious choices are LORAN-C and LF time stations such as WWVB.
> If you are close to a LORAN station but far from WWVB, you will get
> far better results from LORAN.  The propagation differences between
> sky wave and ground wave are like night and day (pun intended).
> You can see the excellent timing performance of LORAN (and GPS) at
> http://phk.freebsd.dk/timedata/
>
> I have recently collected a lot of data from WWVB, and have concluded
> that, at least at my QTH in Reno, NV, it would be very difficult to
> make a good frequency standard which used it.  Before you all say that
> I must have made a major error because the signal could not be that
> bad, please note these quotes from HP's excellent application note
> "The Science of Timekeeping", available for download at
> http://www.tm.agilent.com/tmo/Notes/English/5965-7984E.html .
>
> "Frequency calibrations of local oscillators may be performed by
> continuously monitoring the phase difference between the local
> oscillator and the received LF broadcast. Proper evaluation of the
> resulting phase recordings, however, requires some operator skill
> and experience in interpreting and accounting for various phase
> shifts and possible 'cycle slips'."
>
> "Destructive interference may occur between the first-hop
> skywave and the groundwave, causing a sharp drop in received
> field intensity at certain distances from the transmitter. For a
> 60 kHz LF broadcast, this distance is about 1200 km."
>
> My home is about 1263 km from the WWVB tower :-(
>
> I made a 16 turn balanced loop about 4 by 5.5 feet.  I was quite
> surprised that when the loop was resonated to 60 kHz, one could
> usually clearly see the WWVB modulation on a scope, without any other
> amplification or selectivity.  It was typically about 5 mV p-p.  785
> miles without even using a radio!  I thought that the signal would be
> truly wonderful.
>
> The receiver is very simple.  There is a Lyle Koehler balanced loop
> preamp ( http://www.computerpro.com/~lyle/bal-pre/bal-pre.htm ), a
> JFET mixer, a low pass filter, and a one stage audio amp which feeds
> the line input of a sound card.  The 10 MHz Rb oscillator feeds a CMOS
> counter chip.  The ninth stage (19531.25 Hz) is loosely coupled to the
> mixer LO transformer, which is tuned to the third harmonic at 58593.75
> Hz, giving an IF of 1406.25 Hz.  The 13th divider stage at 1220.703125
> Hz feeds another LPF and drives the other sound channel as a
> reference.  Audio is sampled 8-bit stereo at 11025 Hz; the data is
> processed by a simple C program running under Linux.  Once per second,
> the average carrier amplitude and phase are computed and written to a
> file.  The output is post processed by a perl script and plotted with
> gnuplot.
>
> You can see the results at http://www.scgroup.com/lowfer/wwvb.gif .
> The X axis is in hours, starting at midnight local time (PST) on
> 10-Feb-2000.  The blue trace shows relative signal strength in dB,
> using the scale on the left.  As expected, the signal is generally
> stable during the day and erratic at night.  There is a big dip each
> day centered on local noon.  I am pretty sure that this is the
> destructive interference mentioned above.  An absolute field strength
> scale would make it easier to verify, but I can't think of a good way
> to get one.  The receiver could be calibrated with a known voltage
> input, a preamp bandwidth measurement could establish in-circuit loop
> Q, and the antenna radiation resistance could be determined by
> RJELOOP3.  But the total error when combining these factors could
> easily be 5 dB or more, so the scale would not be trustworthy.
>
> The (relative) signal at night is weaker than I expected.  I am sure
> that there is no significant compression in the receiver, and there is
> no AGC anywhere in this system.  Any explanation?
>
> The red trace shows the signal phase in cycles, relative to the start
> of the experiment, using the scale on the right.  Positive numbers
> represent phase advance.  You can clearly see the hourly pips which
> are caused by WWVB advancing their carrier phase by 45 degrees at 10
> minutes past each hour, and returning to normal phase 5 minutes later.
> I'm quite confident about the phase values; most of the shifts measure
> between 44.5 and 45.5 degrees.  Also, you can physically pick up the
> loop, rotate it by half a turn and see the phase change by 180 +/- 2
> degrees, which is way cool.
>
> There are some obvious cycle slips.  The daytime phase is reasonably
> constant, so the clock can't be way off frequency.  The jump between
> the 11th and 12th must involve at least one slip, and there is clearly
> one just before noon on the 14th.  Unfortunately, I can't reliably
> tell what happens at night, and believe that the nighttime delay is
> actually more than one cycle greater than the daytime.  But I made
> some arbitrary edits, to get the daytime values to roughly line up.
> The result is at http://www.scgroup.com/lowfer/wwvbm.gif .  Now the
> drift in my local clock is evident; towards the end it was gaining
> about 4 microseconds per day.
>
> I can't think of a way to reliably remove the nighttime slips
> automatically.  One could use just the daytime readings to discipline
> an oscillator, but you would have to start with one which is stable to
> less than 1/2 cycle per day.
>
> About once a day, there are sudden phase shifts, sometimes as much as
> 30 degrees in two seconds, and taking a few minutes to return to the
> "normal" curve.  What might cause these?  Meteors?  Solar flares?
>
> Can one use the time code modulation to disambiguate the cycle count?
> This would be very difficult, because of the wide disparity between
> the 60 kHz carrier and the 1 Hz modulation.  I have never seen any
> documentation about the bandwidth occupied by WWVB, so I attempted to
> measure it by looking at the IF signal with Cool Edit.  The rise or
> fall time is about ten cycles of the 1406.25 Hz signal, which works
> out to about +/- 50 Hz.  But over such a short interval, there is
> quite a bit of noise, and each measurement might have an uncertainty
> of about two milliseconds.  Even when averaged over 12 hours (86400
> readings), you would only know the time to within 7 microseconds.
> That's an error of almost half a cycle, so you couldn't reliably tell
> the cycle count.  For comparison, commercial WWVB receivers typically
> specify 100 microsecond accuracy.
>
> At my QTH, LORAN-C is a much different story.  According to the
> directory at http://www.megapulse.com/table.html , the Fallon, NV
> station (400 kW) is only 53 miles away!  My guess is that a trivial
> receiver (one which does not attempt to resolve individual carrier
> cycles, but just measures the average phase of complete pulses) would
> work fine, because the ground wave would overpower sky wave by 20 dB
> or more.  All you need is to divide the local 10 MHz signal by 100,
> feed two mixers in quadrature, and couple the mixer outputs into
> stereo sound card inputs.  The rest is done in software.  If this
> seems workable, perhaps I'll build one and see how it performs.
>
> I started this project, not because of an interest in frequency
> standards, but to gain a better understanding of LF propagation, for
> use in designing a robust ultra-low-bandwidth lowfer signal format.
> WWVB was chosen simply because it is strong and stable.  If the
> propagation mechanisms are the same, these results should be
> applicable to the lowfer band, by simply multiplying the rates of
> change by about 3 to account for the frequency difference.  However,
> it would obviously better to use a signal near 180 kHz.  Can someone
> please suggest one?  I hope to eventually make this measurement on
> TEXAS, but must first find a quiet location for the receiver :-)
>
> Thanks for taking the time to read this long post.
>
> 73,
>
> Stewart Nelson  KK7KA
>
> P.S.
>
> I get "user unknown" when posting to lowfer at qth.net .  Anyone else having this
> problem?  I'm posting to lowfer at ns3.qth.net as a workaround.
>
> To unsubscribe, send to MAJORDOMO at qth.net "unsubscribe lowfer" (Do not
> send to list!!) Send on list submissions to lowfer at qth.net