[ntp:questions] Re: WWVB DSP decoding (like audio refclocks)?

Tim Shoppa shoppa at trailing-edge.com
Thu Nov 6 13:25:04 UTC 2003

"David L. Mills" <mills at udel.edu> wrote in message news:<3FA95802.B5C2365C at udel.edu>...
> High-end WWVB receivers like the Spectracom Netclock/2 use a narrowband
> crystal filter with bandwidth a few hundred Hz ahead of a gain block and
> synchronous demodulator. The I demodulator I channel goes to a slicer
> and computer; the Q channel goes to a 60-Hz VCO. All very expensive and
> close to theoretical optimum.

Thanks, I now know more about what's inside those than I did before!  It's
good to know that they are more than the WWVB-receiver-on-a-chip things
that are in the quote-unquote "atomic clocks".

(Incidentally, for a good response to all the ads for "atomic clocks"
and "atomic watches", see

for the Real Deal.)

> As some folks on this list might remember, the issue has come up about
> WWVB receptino here on the right coast. Conditions seem to vary somewhat
> in various cities, but here in radio-free Newark, DE, conditions are
> awful. My two Spectracom 8170 WWVB receivers on rackety.udel.edu (take a
> look) are 22 years old and have the circuitry used today, but they are
> essentially worthless most of the time. My Netclock/2 at home in a
> noise-quiet location does somewhat better, but still not the rock one
> would expect.
> You hit the nail with your comment on oscillator stability, which is the
> once and for all gotcha. Even with an ovenized oscillator with 10^-8
> stability, synchronous operation could not be assured beyond about 1000
> seconds. You would need an oscillator some 80 times better than that for
> holdover up to a day. This would be in the atomic class.

A disciplined HP 10811E OCXO will run for a day with less than 5 microseconds
of drift, after becoming "unlocked", as in a Z3801A.  That's 6x10^-11
by my calculations.  But the disciplining takes several continuous days
of lock to GPS, and without GPS disciplining the spec is indeed just
a little better than 10^-8.  Those of us on the East Coast are stuck
with *maybe* a few hours each night of clean WWVB reception, probably not
good enough to achieve that level of discipline.

> But, during periods where the signal is above the noise with a good
> antenna, crystal filter and preamp, one of the modern floating-point DSP
> chips like the TI 320-class would work great. Alternatively consider
> using the same general lineup as the Spectracom, but don't use a slicer.
> Offset the VCO by 100 Hz, so now you have moved the carrier to a
> subcarrier within the soundcard passband. The signal is now very similar
> to the WWV/H demodulator/decoder and the same algorithms can be used.
> All you would have to do is change the seconds state machine table.
> Here's a way to test the idea without building anything. Several
> shortwave receivers today can tune 60 kHz with 1-Hz resolution. Wind an
> antenna on a ferrite rode maybe with a preamp. Radios I know about have
> atrocious low gain at and below the broadcast band. Set the BFO to
> produce 100-Hz note and connect your soundcard. Rip off the WWV/H
> driver, toss out the 1000-Hz stuff and synchronize directly to the
> subcarrier. Weekend project.

That's a good and fairly painless idea.  You are correct that most
modern digital radios have very little sensitivity at 60kHz.

> The crystal filter can be a drag, unless you can find suitable rock.

60kHz rocks are readily available, about a buck each, and presumably would
be part of the TRF antenna/preamp.


More information about the questions mailing list