[time-nuts] DC Voltage Ramp?
Lutwak at Alum.mit.edu
Sat Sep 3 19:23:18 EDT 2005
Any decent ovenized oscillator, should deliver stability of 10e-12,
typically flat from 1-100 seconds, but likely getting worse outside of 100
seconds due to drift. A really good oscillator, like the dewarized (??)
1000B may hold 10^(-12) out to 1000 seconds (but I wouldn't count on it).
With a one-way code receiver, the best stability you can get out of GPS is
about 10^(-8)/tau, corresponding to 10 ns jitter at 1 second. As such, it
takes about 10^(4) seconds for it to integrate down to the 1e-12 level, by
which time your OCXO has wandered above it. As such, it is tough to take
full advantage of a good OCXO by steering to GPS.
Standard-performance cesium, on the other hand, typically delivers
1e-11/sqrt(tau), i.e. it integrates down to the level of your super OCXO in
100 seconds (1e-11/sqrt(100) = 1e-12).
For this reason, you will never get "cesium class" stability out of any
OCXO, no matter how good, by locking it to GPS.
Note that you can do pretty well by locking a rubidium clock to GPS. In
this case, you want to set your loop tau of your phase lock to the crossover
of the rubidium stability with that of GPS (about 3000 seconds). Typically
the phase noise of rubidium is pretty poor, so you may want to use your
1000B as a clean-up oscillator by phase-locking it to the rubidium with a
shorter tau, maybe 100 seconds.
After all this, you may ask why the loop tau of the OCXO in a cesium clock
is set to 1-2 seconds, rather than 100 seconds, to take advantage of its
better stability at tau<100 seconds. The reason is that the quartz may
experience a frequency "hop," perhaps exacerbated by vibration or shock to
the instrument, and most users don't want this to persist for several
hundred seconds. In a nice quiet environment, like a standards lab, with a
well-aged oscillator, some users take the risk, turn down the loop gain, and
enjoy the better stability of the OCXO at shorter tau.
Of course, the best of all worlds is cesium locked to GPS (with a loop tau
of a week or more). This was once available as an option board to the
TrueTime XLdc, which steered HP5071 to GPS, or as a package from Datum,
using the 9390 GPS receiver coupled with the 6801 tracking servo and a 4065
or 4040 cesium instrument. There was even a short-lived all-in-one-box
Datum solution which was called Cesium+ and later GPS+. All of these
off-the-shelf GPS-steered cesium solutions are now obsolete, but may appear
on Ebay occasionally.
Robert Lutwak, Senior Scientist
Symmetricom - Technology Realization Center
34 Tozer Rd.
Beverly, MA 01915
(978) 232-1461 Voice RLutwak at Symmetricom.com (Business)
(978) 927-4099 FAX Lutwak at Alum.MIT.edu (Personal)
(339) 927-7896 Mobile
----- Original Message -----
From: "Magnus Danielson" <cfmd at bredband.net>
To: <time-nuts at febo.com>; <brooke at pacific.net>
Sent: Saturday, September 03, 2005 6:48 PM
Subject: Re: [time-nuts] DC Voltage Ramp?
> From: Brooke Clarke <brooke at pacific.net>
> Subject: [time-nuts] DC Voltage Ramp?
> Date: Sat, 03 Sep 2005 15:30:10 -0700
> Message-ID: <431A23F2.9010503 at pacific.net>
> Hi Brooke,
>> The actual oscillator that provides the output from a Cesium standard is
>> in fact a crystal whose fine tuning voltage has been locked to the
>> Cesium source.
>> If the oscillator was run standalone and it's aging rate determined by
>> using GPS then if the proper voltage ramp was applied the output would
>> be very close to a Cesium source. This may be a linear or parabolic
> It would be a bit of both actually. But not quite, the aging is not as
> that it is fully wrapped inside the model of a fixed drift rate. Proper
> stabilizing of the crystal temperature simplifies the model alot thought,
> since shift in temperature would not have the crystal beging a new aging
> period. Infact, you don't want to drive the crystal until you have it
> stabilized on temperature to avoid drift. Some measurement apparatous
> will not turn on until the crystal heating has been on for sufficient
> when it has only then the power button works.
>> An example is plotted at: http://www.rt66.com/%7Eshera/index_fs.htm
>> but for Cesium quality the tuning voltage needs orders of magnitude
>> smaller steps.
>> The problem is coming up with the voltage ramp. I hear that the Austron
>> 2010B Disciplined Oscillator has steps that are too crude. Has anyone
>> come up with a viable way to do this?
> Have you read about HPs SmartClock technology? If not, this is a good time
> do that.
> The HP Z3801A contains the SmartClock technology, and it sure monitors
> alot of things. Mine seems to fail to converge into tighter specs thought.
> In a simple model, you compensate for frequency error (linear phase ramp)
> frequency drift (parabolic phase ramp). In a little more complex model,
> compensate for the temperature dependence. In an even more complex model,
> compensate for the aging process, the magnetic field etc.
> time-nuts mailing list
> time-nuts at febo.com
More information about the time-nuts