[time-nuts] Allan deviation vs standard deviation
jmiles at pop.net
Wed Sep 7 21:26:18 EDT 2011
> Now I vaguely remember: the buffer circuit for the external reference
> is a poor design, because the designers didn't know any better and
> didn't ask for help. (I knew the designers). So even though I am using
> the 10811 inside of an 8662A as an external reference, the 53132
> won't get the full performance of it.
That's not going to happen in any event, because a (good) 10811 is stable to
within 2E-12 or so at 1 second. See attached... if you measured this 10811
on a 5370B you would not get any meaningful data at taus less than about 20
to 30 seconds.
A 5370B is about 6x better than a 53132A, so I scaled the real-world
residual floor measurement from my 5370B by 6 to approximate what you'd get
if you took a residual plot from one (green trace). Here you can see that
you wouldn't be able to characterize this particular 10811 at timescales
shorter than a minute or so. As Said says, anything below about 100
seconds is going to be hard to infer accurately.
The buffer amp might or might not be a problem, but the 150 picosecond
resolution would definitely get in the way unless you enable TI averaging.
The averaging would need to be done in a burst of measurements that's
significantly shorter than the tau-zero period, to avoid biasing the ADEV
plot excessively. The 5370B can do an OK job at N=100 under some
conditions, but I've never tried it with a 53132A.
> I hope the 53230 is better.
> The counters are designed and made in Asia now, so I can't
> exercise the proverbial HP "next bench" paradigm very well when
> the bench is on another continent.
It's supposed to be better. From what the spec sheets indicate, it's as
good as a 5370. Looks like a nice instrument but I haven't seen one in the
-- john, KE5FX
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 60736 bytes
Desc: not available
More information about the time-nuts