[time-nuts] Leap Second results: cheap GPS/1PPS receivers

Chris Caudle chris at chriscaudle.org
Wed Jul 1 16:33:31 EDT 2015


On Wed, July 1, 2015 12:55 pm, Bob Albert via time-nuts wrote:
> I am missing something here.  How does one use a 1 PPS signal?
> I see how I can use a 1 MHz or 10 MHz for a time standard but
> the 1 PPS usage eludes me.

Actually you have it backwards.  A 1MHz or 10MHz can be a frequency
standard, but how do you use it as a time standard?  Time implies a clock
(time of day) reference, and with just e.g. 10MHz you have 10 million
edges per second, you do not know which edge delineates the beginning of a
second, so you cannot use that 10MHz to derive time.

> Unless the pulse is extremely sharp, a minor uncertainty in the
> shape or amplitude will have profound effects on the timing.

The pulse is sharp, but there are other noise sources present on the PPS
output from GPS, so you need other tricks such as averaging to make it
useful for more than just delineating the second in an approximate sense.

> I use an HP 5328A with the OCXO and wonder how I can
> improve its accuracy.

The typical way GPS is used is by using the PPS to (effectively) gate a
counter of a 10MHz source, and the long term average count is used to
steer the 10MHz oscillator so that on average there are 10 million edges
between every PPS.  As I pointed out the PPS from GPS is kind of noisy (in
the time sense, it moves around a bit and isn't actually exactly 1 second
apart each time) so you need a lot of averaging.  That is what takes place
in a GPS disciplined oscillator.
The 10MHz output from the GPS disciplined oscillator can then be used as
an external reference to the 5328A.

If you only have a GPS receiver, and not a full GPS disciplined
oscillator, then you can have the 5328 counter measure the frequency of
the PPS output, and average that over a long period of time.  The long
term average will be 1 Hz, so the difference from 1 Hz is the offset of
your counter time base from the GPS referenced accurate frequency
standard.  Again, you would need to take the long term average frequency
because there are things such as receiver inaccuracies and inherent
atmospheric noise sources which cause small period variations in the time
between each PPS output, but those "should" be random and cancel out with
averaging.

-- 
Chris Caudle




More information about the time-nuts mailing list