[time-nuts] Measuring ADEV with a PM6681 or CNT-81
bruce.griffiths at xtra.co.nz
Fri Oct 24 12:43:48 UTC 2008
Heinzmann, Stefan (ALC NetworX GmbH) wrote:
> So you are proposing to do a kind of simulated time stamping by using a reference PPS signal and measure the time distance between the reference edge and the edge of the DUT? Nice idea, I'd probably need to do some math on the result to convert it to a proper timestamp, but that shouldn't be too hard. Are there tools that can do this for you? Ulrich Bangert's Plotter maybe?
> The limitations of this technique, as far as I see it, are the maximum relative phase deviation of the signal under test with respect to the reference, and the limitations of the counter in terms of dead-time and trigger order. I suppose it becomes much more difficult with higher frequencies.
> The PM6681 does not appear to have support for measuring time intervals in positive and negative directions at the same time. That would limit the usefulness of this trick, it seems. Or is it just as valid to use the opposite edge of one of the signals?
> Heinzmann, Stefan (ALC NetworX GmbH) wrote:
>> Thanks, Tom!
>> Apart from the specifics with TimeView, how does one go about measuring ADEV with a counter that does have a dead-time? Are there
>> tricks that can be played, either with a special measurement setup or with some data postprocessing? Or with two counters that
>> are somehow made to cooperate?
> Consider when you compare two PPS signals. If you counter has a dead
> time of say 100 ms then it is only a certain phase relationship which
> causes the counter to miss a count if you have a counter requiring
> channel A to trigger prior to channel B (or vice versa). If you have +/-
> TI mode then channel A or B can act as a start, and then the other
> channel acts like stop. Then you can hide the dead period as it
> dynamically shifts over.
> The trouble with dead time is much more annoying when you need to count
> the cycles of the signals. The PPS case does that counting without
> having dead-time, so we only need to handle the wrappings of those counters.
>> Sorry if I'm asking the obvious, you can see that I'm a greenhorn. ;-)
> It is a fair question.
> Besides, if Tom sent the question to Staffan, then it will be a little
> while since he is on vacation.
> Björn has a PM6681, he can actually TRY IT. :)
Why not just use Greenhall's picket fence technique which was devised to
allow timestamping of a signal despite the time interval counter having
a finite deadtime?
Use the leading edge of the PPS signal to start the time interval
counter and stop the time interval counter on the next zero crossing of
say a 10Hz, 100Hz, or even a 1KHz signal derived from the local
Since the largest time interval to be measured is 100ms, 10ms or 1ms
depending on the STOP channel frequency, counter deadtime is not a problem.
Details are given in the paper:
More information about the time-nuts