[time-nuts] pps vs. 10 MHz timing

Tom Van Baak tvb at LeapSecond.com
Fri Dec 12 03:53:36 UTC 2008


Hi Matt,

> What do other GPSDOs do?

All over the map. A well-engineered GPSDO might choose to
align the rising zero crossing of the 10 MHz sinewave with the
leading edge of a very fast risetime 1PPS, because this is
"common sense" but if it isn't in the spec, it's best not to expect
it to happen.

> Clearly, any device trying to latch the 1 PPS signal using the 10 MHz
> clock will need to choose which clock edge to use, depending on which
> type of GPSDO is used.  How is this normally handled?

But worst case, if you chose the wrong 10 MHz sinewave edge
you're at most 50 ns off from the 1PPS -- and note this is still
within the +/- 100 ns of UTC accuracy spec that a lot of GPS
receivers claim.

There is often a fixed phase delay between the two outputs. It
may be just a few ns or it may be tens of ns depending on the
number of inverters, buffers, amplifiers in the signal path(s),
the length of internal or external cables, or impedance effects
on risetime (50R vs 1M can make a difference).

One also has to be very careful with the 1PPS pulse itself. Most
GPSDO do not specify the voltage level of the pulse. Now if the
pulse risetime is just a couple of ns this is not a serious problem.
But when the risetime is many ns or many tens of ns then you
need to know what trigger voltage level constitutes the intended
true moment of the 1PPS. See, for example, 

http://www.leapsecond.com/pages/gpsdo-rise/

Some of these GPSDO have about a 1 V / 10 ns slew rate which
means if you don't nail down your trigger level you're asking for
very large shifts in 1PPS timing. The zero-crossing of a 10 MHz
sinewave doesn't have the same ambiguity.

/tvb




More information about the time-nuts mailing list