[time-nuts] means of calibrating/verifying Ru with GPS
bruce.griffiths at xtra.co.nz
Sat Dec 22 05:34:43 EST 2007
> 1E7 seconds eh?
> Well, maybe I don't need that degree of accuracy after all. Thanks for
> the quick response. How are these Ru oscillators calibrated in a
> commercial environment? what's the likely frequency tolerance on my
> second hand Ru osc? I need to revise my theory I think - it been a
> while since I opened the maths books. How do the GPSDO deal with this
> or is the stability good enough for them? What can I do to improve my
> confidence in the frequency output of the Ru? I would like to use this
> Ru as a house standard for all my frequency counters etc.
A GPSDO in effect averages the measured PPS phase error over a long time
(hundreds to thousands of seconds depending on the stability of the
oscillator being disciplined).
In other words the Phase locked loop bandwidth is very low.
With a good timing receiver (eg M12+T, M12M T) and antenna setup, the
PPS rms timing error after sawtooth correction may be as low as a few
If the GPS receiver PPS quantisation error is 1us then a gate time of
around 10,000 sec is required to reduce the measurement error to about
If the rms error is 10ns then a gate time of 100sec will suffice.
One of the better ways to improve your confidence in tRu oscillator
characteristics is to time stamp the PPS signal from a GPS timing
receiver using a timebase derived from your Ru Oscillator.
The easiest way to do this is to use the 5335A to measure the delay from
the leading edge of the PPS pulse to the next leading edge of say a
100Hz clock (produced by dividing down the Ru oscillator output. This is
known as the picket fence technique. If the measured time delays are
logged for several hours then algorithms exist to unambiguously unravel
the measurements and convert them to time stamps which can then be
processed to derive parameters such as frequency offset etc.
More information about the time-nuts