[time-nuts] Allan Variance with an HP53132a counter
magnus at rubidium.dyndns.org
Sun Jan 22 14:34:32 EST 2017
There is two types of errors done when using this type of counters,
First, the delta-estimator of frequency filtering will skew the Allan
deviation, mostly in the white phase noise region but the effect wears
of at the length of the averaging windown of the counter. Correct
interleave factor and processing lets the prefiltering be extended into
modified Allan deviation.
Second, the moving average is a form of interleaved estimation producing
a higher reading rate than the length of the avereage, giving a improved
response compared to traditional non-interleaved behavior. However,
processing such values as non-interleaved values will skew the Allan
deviation response. Correctly handle it as interleaved values removes
this bias effect.
You can do the same with Omega counters, considering you do things
properly for PDEV. I've got a paper to complete on that topic.
On 01/22/2017 03:52 PM, Gilles Clement wrote:
> I am new to this list and have been following it for a while with great interest. I am afraid I am getting contaminated with the "TN" virus… !
> I developed a home-brewed OCXO crystal oscillator disciplined with a long wave radio station signal (162 kHz). Its working quite well with a long term stability that is « probably » better than 10E-8 which I am actually looking forward to better understand. I am considering using an HP53132 counter for this purpose, although I understand that it may not be the most recommended approach….
> Please find hereafter the rationale:
> - The HP53132 implements a triangular averaging algorithm (Delta averaging) in its standard frequency or period measurement mode of operation
> - This feature provides a pretty high resolution (up to 0.5*10E-12 if I am correct) which should be appropriate at least to get a first idea of the oscillator performance
> - But it comes with the drawback that an Allan Variance computed directly from such frequency (or period) readings, would be distorted with respect to the "standard" Allan Variance (which assumes a PI averaging instead of a Delta averaging algorithm to compute the fractional frequencies)
> - However, from the Australian paper titled : « Considerations on the Measurement of the Stability of Oscillators with Frequency Counters » I noticed that:
> When I look at OCXO crystals oscillator stability discussions in the literature, White Frequency, Flicker Frequency and Random Walk Frequency seem to be the dominant noise factors considered (ex: in phase noise spectral density graphics)
> The distortion from Allan Variance to Delta Variance is limited for White Frequency, Flicker Frequency and Random Walk Frequency noises (though it can be several orders of magnitudes for White Phase and Flicker Phase noises)
> Moreover formulas are provided in the paper to evaluate the amount of such distortions. The impact appears to be a simple multiplier of 1.33, 1.30 and 1.15 for the three corresponding noises, the slopes of the curves are not impacted (i.e: tauE-1 ; tauE+0 and tauE+1)
> - So the idea is the following: start with a simple bench test, reading the standard mode of Periods measurement, calculate the "naive" Triangular Variance from these data, identifying the noise types (according to the slope of the curve sections), and applying the corresponding offset correction, hopefully leading to an estimate of the "standard" Allan Variance.
> - This approach would have the great advantage to be simple and cost effective, as one can find nowadays second hand 53132’s at reasonable cost even with the (mandatory) OCXO option.
> Comments welcomed,
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
More information about the time-nuts