[time-nuts] Oops - wrong URL for tvb plot!

Tom Van Baak tvb at LeapSecond.com
Tue Apr 10 21:43:22 EDT 2007


Said,

The nice thing about 1PPS measurements is that they are
known to be regular, in spite of delays in the counter or the
measurement system. For mid- or long-term analysis one
sample a second is more than enough. For example, if you're
making a tau 1 second to tau 1 day ADEV plot you don't
need more data than this.

Using 10 MHz as input works too but, yes, now the data
rate is more dependent on the internal delays of the counter.
Note also for tau less than 1, the internal noise of the counter
is likely to exceed the noise of either of your sources, so
the data you get can be misleading.

Knowing or calculating the sample rate is useful: 15-20
samples/second seems OK for a 53132. But if you were
to measure it you would come up with a more exact number.

One way to do this is time the collection of 1000 samples.

The way I do it is by time-stamping all the samples. You're
welcome to use my comlog.exe tool. It's hiding under:
    http://www.leapsecond.com/tools/
Use the /ms option or the /mjd options.

Then, knowing the exact sample rate gives you the base
tau for all further calculations.

On the 53132A you can also specify a gate time. Here
you get fewer samples per second, but each is more
accurate (by averaging).

The last idea is the safest: use an external gate as you
suggested. You will get 10x more data; but remember
that for all this extra 10x more data all you get is one little
data point; tau 0.1 seconds.

It is sometimes simpler to measure very short-term
stability (e.g., 0.001 to 1 Hz) with a short test (a minute
is enough) and then collect 1 Hz data for hours or days
for the rest of the tau range.

For the timebase, yes, it could well be that the high-stab
internal timebase has better short-term stability than your
Cs. Hard to predict, but simple to measure: compare the
stddev of your UUT data using the internal OCXO vs. using
the external Cs.

/tvb

----- Original Message ----- 
From: SAIDJACK at aol.com 
To: tvb at leapsecond.com ; cfmd at bredband.net 
Cc: time-nuts at febo.com 
Sent: Tuesday, April 10, 2007 4:37 PM
Subject: Re: [time-nuts] Oops - wrong URL for tvb plot!


In a message dated 4/10/2007 14:48:34 Pacific Daylight Time, tvb at LeapSecond.com writes:
with a 53132A against a good atomic house standard. You'll
see a large phase offset (many microseconds), a frequency
offset (about +3e-12 or -3e-13). No, I didn't remove frequency
drift (if any) from the data before I plotted it.

Hi Tom,

how do you best compare an atomic standard against an OCXO on the 53132A?

I did it with the 1PPS inputs (that works, but gives only one sample point per second).

Inputting 10MHz and doing T1 to T2 intervall measurements also works somewhat, except using the internal gate delay is not perfect, I get aliasing (there is some 100ns offset sometimes - which is random).

It works best (no aliasing) when I let the counter free-run without any gate-delay, but then I do not know the sample Tau of my data.

That's partially because I am using the serial port output at 19200 Baud, and I think the counter is spitting data at me as fast as it can (limited by the RS232 port). Seems its about 15-20 measurements per second free running.

I am thinking that maybe I should provide a 100ms external gate input etc, at least that way I will know for sure that my Tau is 0.1s and I get 10x more data than simply comparing the 1PPS outputs.

Lastly, I have the High Stability time base, but have still been using an external Cs reference to feed 10MHz into the counter. I wonder if the internal time base is better (probably less short term ADEV than the Cs) for these short time intervall measurements.

Any hints are appreciated,
Thanks,
Said


More information about the time-nuts mailing list