Why is HP 5370/5345 Time Interval Measuring less accurate trhan Period Measuring

[Peter, PE1ECM]

I use a HP 5345B and a HP 5370A for time interval measuring ( separate
as well as common mode ).
The 5345A has a jitter stability ( useable resolution ) of 2 nS and
the 5370A a jitter stability of 20 pS.
Both the instruments behave OK.
However, I see that the stability of the period is far better than the
accuracy of the time interval.

Whatever tweaking I have done and whatever signal source I use ( e.g.
a e-11 frequency reference )and at whatever frequency I perform the
measurements, any time interval will be no more stable than about 5e5.
Thus, measuring ( 5370A )( common TI mode ) a 10 Mhz frequency
reference signal the stability shows about +/- 40 pS jitter which is
very good and far within the specs.

Measuring at 1 Mhz ( reference locked ) however, the stability is +/-
some hundreds of pS.
And measuring at 10 Khz , the stability is +/- 50 nS

I get about the same ( relative ) results when using the 5345B instead
of the 5370A ( be it that the jitter stability and resolution of the
5345B is 100 x less ).

However, measuring the period ( instead of the TI mode ) I get the
expected results of about e-11( which is cf the specs if locked to a
better than e-11 freq.ref ).

In the frequency modes both the instruments perfom far within specs (
stability resolution better than e-11 ).

What is it that TI measurement ( in these particular HP gear )is so
much less stable than period measurent ?

Since both the instruments show this very phenomenum and for both even
in the same relationship at varying frequencies, I guess this must be
system feature and no flaw.

Any HP guru in this group ?

I bent my mind for days but I cannot resolve this one.
Thanks for help.

Peter, PE1E ex PE1ECM

 

[Jim Adney]

What is it that TI measurement ( in these particular HP gear )is so
much less stable than period measurent ?

There is a built-in time uncertainty of one internal clock oscillator
period in any counter. This is the resolution of your counter.

For an TI measurement you are measuring a single event, so your
resolution on that measurement is your uncertainty.

For a period measurement, many instruments will measure over N periods
and then divide by N to get the period. The same resolution above
applies to the total time measured, but this makes the uncertainly for
a single period 1/N times the measured uncertainty.

Often you can select N when you set up the instrument; it will be
related to the gate time that you use when you measure frequency. Just
as your frequency measurement gets better if you average over a longer
gate period, so will your period measurement.

-