[time-nuts] new paper on Allan Variance errors

Matt Osborn kc0ukk at msosborn.com
Sat May 19 00:40:21 EDT 2007


On Fri, 18 May 2007 19:23:24 -0700 (PDT), you wrote:

>
>The latest IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency
>Control  (May 2007) has a very important paper by Dawkins, McFerran, and Luiten
>"Considerations on the Measurement of the Stability of Oscillators with
>Frequency Counters"
>
>Many of us use or have used a high resolution reciprocal counter to measure
>short term stability of a source. This paper shows that the result is often off
>by 33% and can be off by more than an order of magnitude, depending on the
>amount of frequency and phase noise in the DUT.
>

Which way?  I'm inexperienced  in these matters and am trying to find
my way.  My goal (for now) is to build a GPSDOCXO, but I don't
understand what it is I see when trying to characterize an OCXO.

I bought a used HP 53131A counter and have started to compare the
stability of two Isotemp OCXOs with the various 'standards' available
to me.   I've been disappointed with the results so far, but then I'm
not sure in which I should place my confidence or even if my
dissapointment is warranted.

The two Isotemp 10MHZ OCXOs  are rock solid at 1/10 hz resolution, but
at 1/100 Hz  I see drift up to +- 12/100Hz within 1/2 second.  I've a
used HP8656B signal generator that drifts less than +- 5/100 Hz in the
same time frame.

When measuring stddev with a gate time of 0.5 second period, I measure
roughly 0.0039 with the Isotemps and .0020 with the HP 5313aA

From this I've concluded that the HP 8656B and the HP 5313A are both
probably an order of magnitude better than either of my Isotemp
oscillators.

The Isotemp oscillators are both running uncorrected at this point and
both are approx. 0.5 Hz lower than their nominal frequency of 10Mhz.

Since the variance I see is rapid (up in one 1/2 second period and
down in the next) I've assumed that what I'm seeing is phase noise.
My assumption is that a frequency stability issue would manifest
itself as a longer term drift in one direction or the other.

I'm assuming the paper you reference advises that our frequency
counter measurements lend more credence to the DUT than is warranted.
However, hope springs eternal...





More information about the time-nuts mailing list