[time-nuts] Picket fence technique

Magnus Danielson cfmd at bredband.net
Sat Dec 23 09:39:02 EST 2006

From: Dr Bruce Griffiths <bruce.griffiths at xtra.co.nz>
Subject: [time-nuts] Picket fence technique
Date: Sat, 23 Dec 2006 18:07:43 +1300
Message-ID: <458CB99F.2040103 at xtra.co.nz>


> The Question is has anyone considered using this technique?

I have not seen it in use for any commercial products. In fact, few of them
up until recently did ADEV and friends builtin. Futher, few manufactures seems
to have a full grasp of measurment tricks around ADEV and some have even
ignored the dead time aspects in their measurements.

Besides, things changed considerable when non-dead time counters became
available. With todays technology, it is not very hard in building non dead
time counters or for that matter buy them of Ebay (such as HP 5371A or 5372A).

One technique of TIC measurement wich may overcome dead time issues but which
would work for ADEV would be to measure three time intervals (or two frequency
measures) back to back. These can be processed and squared separatly with the
correct tau. Doing incremental measures would be dead-simple and you could see
the ADEV measure creep down for each measure taken. That will not work very
well for MDEV and TDEV and you will not be able to compensate for frequency

There is another incremental method for ADEV which will allow for frequency
drift removal. This is done by summing both the squared and un-squared results.
A simple post-processing will be equalent of first scanning the samples to take
the average and then removing that average from the samples before taking the
squared result for the actual ADEV. The necessary processing memory is very
small all processing is done per sample as they arrive.

I haven't looked at the details of MDEV/TDEV and incremental form, but now that
christmas vacation is here, there is something to do.

Merry Christmas and a well timed Good New Year!

More information about the time-nuts mailing list