[time-nuts] Second FTS4060 shows Drift, is it me? Good Links
brooke at pacific.net
Tue Feb 7 14:22:07 EST 2006
No, the plots are of the raw data. Comparing this morning (8:27) @ 892
ns with yesterday morning (8:27) @ 858.2 gives a linear slope of 3.7E-13.
This may be the Ultra Stable Oscillator, but I don't see how changing
the C-field would effect it. That is, if the C-field is changed then
the plots change.
That's an interesting idea about the 1 Mhz. But if there's a cycle slip
it will have no impact since the counter will only show 0 to +1000 ns.
What does have an impact is when the plot crosses 0 or 1000 ns. For
example when the nominal data was just above zero and sloping downwards
for a few days the data had a lot of noise (higher by a few hundred ns)
since there were data points just above zero and just below 1000 ns in
the same 500 second average. But this is easy to recognize. The 500
second jitter is around 9 ns which is just the M12+T sawtooth. Last
year multipath caused higher jitter (hundreds of ns) when the GPS
elevation mask was set near the horizon. Now the El mask is much higher.
If the counter trigger point moved then there would be a problem. Need
to think of some type of test of just the counter.
Thanks for the thoughts,
John Ackermann N8UR wrote:
>Hi Brooke --
>Did you remove any linear offset before generating thorse plots? They
>sure look a lot more like an Xtal aging with offset removed, than a Cs.
>I wonder if it might be related to your use of the 1MHz output, rather
>than 1pps, to drive the stop input. Maybe you're slipping cycles due to
>drift in the counter timebase (or some other systemic reason)? I've
>sometimes seen squirrely results when trying to use a high frequency
>from the DUT to drive the stop signal. Dividing down to 1pps would
>eliminate that as a possible problem.
>Brooke Clarke wrote:
>>Sorry for the bad links, here are good ones:
>>Today's plot = http://www.pacificsites.com/~brooke/pdf/sn1013_850_Drift2.pdf
>>s/n 1227 29 Apr 2005 plot =
>>Brooke Clarke wrote:
>>>A year ago I took apart the FTS4060 that was DOA (s/n 1013) because of
>>>rough shipping and used it to figure out most of the interconnect
>>>wiring. A few months ago I put it back together with the thought of
>>>selling it as a parts unit. But first powered it up and found that it
>>>now locked and seemed to be working so tried to set the C-field.
>>>For the last couple of weeks it has shown a parabolic plot like s/n
>>>1227, although this time the polarity is opposite that of s/n 1227 which
>>>also showed drift, but that may be a setup difference. A plot of s/n
>>>1227 is at: file:///C:/Webdocs/pdf/Cs_Drift0429.pdf
>>>The equation for s/n 1013 is:
>>>y = 2.7943x2 - 302.64x + 8969.4 and the quality of fit is
>>>R2 = 0.9088. The x-axis is in days and the y-axis is in ns.
>>>The first derivative of the equation has a first term of 2 * 2.7943 * x
>>>ns/day or +5.3E-14 seconds/seconds drift rate.
>>>Today's plot is at: file:///C:/Webdocs/pdf/sn1013_850_Drift2.pdf
>>>The current setup is:
>>>SR620 time interval counter doing 500 averages (500 seconds = 8 1/3
>>>Start from Motorola M12+T (9 ns jitter).
>>>Stop from FTS4060 1 MHz output.
>>>Manually enter into spreadsheet date and time of reading (usually not on
>>>the 8 1/3 minute change, just a random time) and counter value.
>>>For most of the test the SR620 was using it's internal oscillator and
>>>just recently I changed it to the PRS10 external standard but that does
>>>not seem to have made any difference.
>>>Is there something I'm doing wrong that would cause apparent drift?
>>>Thanks for any thoughts,
>>>Brooke Clarke, N6GCE
>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