[time-nuts] Subject: Re: Allan Variance with an HP53132a counter

clemgill at club-internet.fr clemgill at club-internet.fr
Mon Jan 23 12:03:03 EST 2017

<span style="font-family:arial,helvetica,sans-serif; font-size:12px">‌</span>Hi Magnus,<br>
Thanks for your enlightening reply.<br>
I am indeed looking at a possible process to correct the DeltaDEV bias to provide something approaching the traditional ADEV.<br>
Could you please send me a link to your paper on Omega counters ?<br>
Indeed the "Australian" paper shows significant skew in White phase and Flicker phase noise regions but much less in the other noise regions, which shoudl be correctable when one identifies the shape of the responses.<br>
Another paper from E. Rubiola ("On the measurement of frequency ans of its sample variance with high-resolution counters") introduces the point that the average on m contiguous Delta measures converges to a Pi estimate (AVAR) for large value of m. However this seems to be challenged by the "Australian" paper pointing at a significant side effect (sensitivity surge) around N times the reciprocal of gate time (N = number of samples). <br>
<div class="gl_quote" style="margin-top: 20px; padding-top: 5px;">
<div class="gl_quoted">Message: 11<br>
Date: Sun, 22 Jan 2017 20:34:32 +0100<br>
From: Magnus Danielson<br>
To: time-nuts at febo.com<br>
Cc: magnus at rubidium.se<br>
Subject: Re: [time-nuts] Allan Variance with an HP53132a counter<br>
Content-Type: text/plain; charset=utf-8; format=flowed<br>
There is two types of errors done when using this type of counters,<br>
First, the delta-estimator of frequency filtering will skew the Allan<br>
deviation, mostly in the white phase noise region but the effect wears<br>
of at the length of the averaging windown of the counter. Correct<br>
interleave factor and processing lets the prefiltering be extended into<br>
modified Allan deviation.<br>
Second, the moving average is a form of interleaved estimation producing<br>
a higher reading rate than the length of the avereage, giving a improved<br>
response compared to traditional non-interleaved behavior. However,<br>
processing such values as non-interleaved values will skew the Allan<br>
deviation response. Correctly handle it as interleaved values removes<br>
this bias effect.<br>
You can do the same with Omega counters, considering you do things<br>
properly for PDEV. I've got a paper to complete on that topic.<br>

More information about the time-nuts mailing list