[time-nuts] Time interval counters and low S/N signals.

Poul-Henning Kamp phk at phk.freebsd.dk
Fri May 20 11:30:08 EDT 2005


In message <428DDE92.2010407 at medphys.ucl.ac.uk>, "Dr. David Kirkby" writes:

>Anyone ever tried this sort of thing? Comments?

Amplitude noise translates directly to time jitter in time-interval
measurement.

There are two sources of jitter:  zero-crossing slope and false
zero-crossings.  As your S/N decreases, the risk of false zero-crossings
increase.

zero-crossing slope results in gaussian time jitter and averages
out well if you can average enough samples.

false zero-crossings are usually stratified on the noise-spectrum
and may or may not average out at all.  In this case you need to
do a histogram and apply relevant statistical tools depending on
the shape.

If you run the counter in 50Ohm mode and with decent voltages, you
don't have to worry about charge-threshold issues, which is a very
hairy ball of vax to get into.  (At low signal levels or high-impedance
mode the trigger point is not only a voltage but also a certain
amount of charge to fill the input capacitance).

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.




More information about the time-nuts mailing list