[time-nuts] Some comments on GPS timing
tom.k3io at gmail.com
Thu May 24 00:43:53 EDT 2018
Perhaps it is time that I step in to point the newer members of
Time-Nuts to some of the historical (hysterical?) info on using GPS to
achieve sub-microsecond global timing sync. I refer you all to the
website http://gpstime.com that Rick Hambly (http://cnssys.com)
maintains to document some of our timing exploits over the past ~25
years starting with the early Motorola PVT-6 "SIXPACK" and proceeding
thru the later ONCORE, M12 and UBLOX receivers.
By explanation, I was one of the principal researchers that developed
the technique of Very Long Baseline Interferometry (VLBI) for high
accuracy Geodesy and Astrometry. In VLBI we record raw noise from a wide
RF bandwidth at a network of radio telescopes around the world. Most of
the noise picked up at the stations is from the microwave receivers, the
earth's atmosphere & the 2.7 degree remnant background noise from the
big bang -- the system noise resulting from all these sources at each
station is independent and uncorrelated with noise at all the other
But a small fraction of the noise originated at quasars which are
(nearly) point sources with an angular sizes measured in fractions of
milliseconds of arc (1 milliarcsec is about the size of George's head on
a quarter held over over San Francisco as seen from Boston); the noise
from these compact sources is correlated at the stations separated
thousands of km and serve as reference points for the celestial
reference frame. The telescopes define fixed points on the earth which
constitute the terrestrial reference frame.
To extract the correlated signal VLBI each station uses an independent
Hydrogen Maser clock as a time and frequency reference. The independent
clocks need to be synchronized at nsec levels and to hold this
synchronization for time scales up to several days. I realized that, if
each station used a GPS-based timing system operating 24/7/365, we could
solve the time/frequency sync needed to measure the "real-time" motions
of the earth as the continents drift and the rotation of the earth and
the sub-milliarcsec astrometric position and structure of the quasars.
FYI -- After 3+ decades of such measurements, we now know the few cm/yr
(about the same speed as your fingernails grow) of relative motions of
the continents (a.k.a. "continental drift") to uncertainties of tens of
microns/yr. The celestial reference frame as defined by the positions
~1000 extragalactic radio sources is known to ~10 microarcseconds. You
can see get a feeling for these results and the network of stations that
produced them by browsing http://ivscc.gsfc.nasa.gov .
As you browse the material on http://gpstime.com you will see snapshots
of our efforts at delivering low-cost GPS timing system intended for the
VLBI station operators ("telescope drivers") updated every couple of
years. Also you will find presentations we gave at several Precise Time
and Time Interval (PTTI) meetings.
To answer a couple of the recent questions in more detail, take a look
at the 2006 "Low Cost" PTTI paper :
1. For all the ONCORE and UBLOX receivers: The "sawtooth correction" is
in the binary data message for the NEXT second. The receivers have
an counter register that is updated based on the navigation firmware
(typically a Kalman filter) to the integer number of clock counts
corresponding to the next 1 PPS epoch. The sawtooth correction is
the fractional part of the epoch that is left over after the integer
clock count is set into the hardware.
2. In Rick's CNSCLOCK, a programmable delay line is fed from the
fractional counter error (the "sawtooth correction" plus a small
constant "DC" to center the correction in the delay line). The "DC"
bias is treated as if it were a cable length correction.
3. My email address has changed from verizon.net: Now it is
mailto:tom.k3io at gmail.com
73 de Tom K3IO
This email has been checked for viruses by Avast antivirus software.
More information about the time-nuts