[time-nuts] Phase microsteppers
brooke at pacific.net
Thu Mar 10 18:10:02 EST 2005
Here is some data taken today. First coul is clock time when I read
counter and second coul is time interval in ns.
I may have missed one or more of the readings. Is this enough to get a
1,000 second Allan variance? If so how does the FTS4060/S24 compare to
other Cesium clocks?
clock time when I read the counter:Tom Van Baak wrote:
>Before I answer the microstepper questions let me
>first describe the alternatives.
>Let's suppose you have the following:
>- a working cesium (or rubidium) standard,
>- a government reference (such as Loran or GPS),
>- time interval counter, and perhaps
>- some sort of PC logging (GPIB, RS-232, etc.)
>Within hours, or certainly days, you notice your Cs
>1 PPS is diverging from a UTC 1 PPS. So what to
>do about this?
>You can compute the average Cs frequency error
>with just two data points since frequency is delta
>phase over delta t. So you really don't even need
>PC logging for this. Just look at your TI counter
>display, wait a day and look at it again. 864 ns a
>day is 1e-11, 86 ns a day is 1e-12, etc.
>Or you can get fancy and plot phase and look at
>the slope. If the slope is ragged, wait longer until you
>see a clear trend. This could be hours or even days.
>If you don't like amount of frequency error you can
>try to adjust the C-field. It will take several tries but
>you should be able to get your Cs accurate to within
>1e-12; maybe better depending on your eBay luck.
>If you want to get closer than this you may want to
>first check the stability of the standard. After several
>days or weeks of data you can compute the Allan
>Deviation of your standard for, say tau 1 day. Even
>just one data point per day for a week is enough to
>give you a rough ADEV result.
>If you find the ADEV doesn't get much below 1e-12
>then there's no point in trying to set the accuracy
>of your Cs below 1e-12. Stability is not affected by
>accuracy but accuracy is limited by stability.
>No clock is perfect. Clocks have noise floors. At the
>1e-12 or 1e-13 level frequency will wander, phase
>error will, over time, grow unbounded. The same is
>true for wristwatches (just move the decimal point
>over far enough).
>The reason I stress computing the ADEV is so that
>you can adjust your expectations. I'm pretty sure no
>amount of C-field adjustment of a standard 5061A will
>get you down to 1e-13 or the 14's.
>All Cs standards, once set, will show accumulated
>phase errors, no matter how carefully you set the
> rate or control the environment. So you should
>expect over days, or weeks, or months, that your
>Cs will be off by many nanoseconds, even many
>It then becomes humbling that a $25 surplus GPS
>receiver will easily beat your cesium as source of
>a UTC tick ;-)
>So the first point I wanted to make is don't expect
>your Cs to ever be perfectly accurate.
>The next question is, after you get the C-field set
>to your liking, how to make further time or rate
>time-nuts mailing list
>time-nuts at febo.com
w/o Java http://www.pacificsites.com/~brooke/PRC68COM.shtml
More information about the time-nuts