[time-nuts] Characterising frequency standards

[Magnus Danielson]

Mark,

Mark Sims skrev:
> The whole purpose of taking a data set from a known ZDT counter and then throwing out random samples is to simulate the kind of data that a normal counter would produce.  You could compare the results and get an idea of how using a normal counter for calculating adevs would compare to using a ZDT counter. I would start by generating random numbers from 0-3 and throwing out that many samples.
> 
> With most normal counters you cannot guarantee that you would get a sample every other interval.  It all depends upon how the counter works,  what its timebase is,  how it triggers and retriggers,  how it is being read out,  what the input signal is, etc.   I would suspect that most counters would give a reading every two or three intervals.  I have seen some counters give two consecutive back-to-back readings then a long dead time. 

Most counters I know of would make one frequency measure, then miss the 
directly following just to trigger directly ontop the next, those for a 
PPS pulse it would measure the period between the first and second 
pulse, then dwell until the third pps pulse and measure until the fourth 
pulse, but then happily repeat this pattern.

But measuring frequency/period like this is not very useful for 
post-processing in any Allan Deviation measure. The lack of back-to-back 
measures prohibits you from achieving the data you need.

We rather use time-interval measures. Let's consider the same counter, 
we arm it with a PPS pulse from either of the sources, but then measure 
the time interval between two 1 kHz variants of the signal, or use the 
PPS as start of the TI and the stop channel sees the 1 kHz signal, I'll 
use the later as a reference, but the cases are equalent.

The same counter can now dwell between the measurements, but most 
counters can withstand 1 measurement per second without too much 
trouble. The 1 kHz signal allow for a maximum of 1 ms delay from 
arming/start trigger to stop trigger. This still allows for plenty of 
time for the counter post-processing to occur and re-arming. As the 
clocks drift, dynamically would stop-channels choice of 1 kHz flanks 
shift, but it would be a fairly simple task to post-process that into a 
continous stream of PPS marks.

Using these time-interval measures of tau0 being 1 s, we can now make 
any set of back-to-back frequency measures as we please, as long as they 
are integer multiples of tau0 by dropping n-1 samples inbetween and 
recall that the sample-series has converted to a tau0 of n seconds.
We can also use the series directly for the Allan Deviation estimator of 
choice in either time or frequency form.

Thus, the lack of zero dead time does not necesserilly prohibits the 
use, but care in setting up the signals and I/O can curcumvent the problem.

Many counters is being used one way or another for continuous measures 
even if they are not exclusive ZDT counters, but it takes care.

Having one or two of TVBs PIC dividers at hand should certainly be handy 
for doing tricks like this.

Time-resolution of the counters as well as trigger noise may be issues 
to look at.

When do one need true ZDT counters then? Well, if you want to make 
measurements for higher frequency modulations, you need that power, but 
most of the time they are just very handy tools.

Cheers,
Magnus