I have several of these HP ovenised 10MHz oscillator modules which are
not able to be run 24/24 although I leave the ovens operating (they
are generally in power-hungry HP test equipment which has a standby
mode which leaves power to the oven).
I've been noticing that they take a decent number of hours (minimum
8h) of running as oscillators from standby before the short term
frequency is adequately stable (in terms of not drifting) for some
purposes. It's very obvious when trying to phase lock multiples of two of them
together using EFC to tune one and a narrow bandwidth PLL. In trying
to find zero beat the beat frequency can be seen drifting on a scope /
counter fluctuations usually around 0.5-1Hz over 20-30min period.
During this time they compare less well to a boring modern non-ovenised
oscillator for short term stability. No actual quantitative
information because there isn't a better standard in this location to
use as a timebase, although I do have an HP53132A counter (but with
the rubbish timebase option) which is what I use to measure the
beatnote frequency.
I'm wondering what the physical process might be for this settling? Is
heat dissipation in the oscillating crystal needing to find thermal
equilibrium? Or do they need repair in some way? I've never taken one
apart but I do have the service manual.
The manual for the 10811A suggests running for 10days before checking
drift, but to do so requires a Caesium frequency reference which -
alas - I lack!
I'm aware that there's a lot of knowledge of these HP oscillators
around here and would be grateful for any pointers.
Alan
Hi
Your OCXO (pretty much any OCXO) goes from room to “pretty hot” ( that might be 20C up to 90C)
in a short period of time. The temperature coefficients of expansion of the various elements of the
crystal never match up perfectly. The same is true for the “stuff” in most of the other components.
This stress translates to a change in value for the components and a change in frequency for the crystal.
As time goes on, this stress relaxes. If you cooled the part(s) down and still could observe the values,
you would see a change from that as well. It also would eventually relax out.
Regardless of how well the crystal is scrubbed, there always will be “stuff” inside the crystal enclosure.
It will settle into some sort of equilibrium over time. Change the temperature and the equilibrium point
changes. That can take a long while to get to some sort of near stable state.
The typical 10811 is not a sealed unit. Humidity (and dirt) can get into the package. If the unit has been
in storage for a long time, it can take a while for the boards to bake out / crud to burn off. Until it does, the
unit really isn’t “in good condition”. Spending years in a damp and dirty garage might mean months of power
on to get things to “fly right”.
Are there a lot of other contributors? Sure there are. You could fill a pretty good sized bookshelf with
all those details.
The 10811 should change about 1x10^-9 from 1 hour on to 24 hours if it is in good condition. If you are
multiplying up to 1 GHz, that would be a 1Hz change. If you are looking at 10 MHz, a 1 Hz change is 0.1
ppm. A properly functioning 10811 should easily hold 0.1 ppm per hour after 1 hour on power.
If you are looking at phase rather than frequency, that’s a different thing. A 360 degree change in one
second gets you 1 Hz. A 360 degree change in 20 minutes is 1/20 th of a hertz. That gets you to a
frequency offset of 5x10^-9. If you are looking at the the first hour on power, that’s not an unreasonable
number.
Bob
On Mar 31, 2024, at 10:59 AM, alan bain via time-nuts time-nuts@lists.febo.com wrote:
I have several of these HP ovenised 10MHz oscillator modules which are
not able to be run 24/24 although I leave the ovens operating (they
are generally in power-hungry HP test equipment which has a standby
mode which leaves power to the oven).
I've been noticing that they take a decent number of hours (minimum
8h) of running as oscillators from standby before the short term
frequency is adequately stable (in terms of not drifting) for some
purposes. It's very obvious when trying to phase lock multiples of two of them
together using EFC to tune one and a narrow bandwidth PLL. In trying
to find zero beat the beat frequency can be seen drifting on a scope /
counter fluctuations usually around 0.5-1Hz over 20-30min period.
During this time they compare less well to a boring modern non-ovenised
oscillator for short term stability. No actual quantitative
information because there isn't a better standard in this location to
use as a timebase, although I do have an HP53132A counter (but with
the rubbish timebase option) which is what I use to measure the
beatnote frequency.
I'm wondering what the physical process might be for this settling? Is
heat dissipation in the oscillating crystal needing to find thermal
equilibrium? Or do they need repair in some way? I've never taken one
apart but I do have the service manual.
The manual for the 10811A suggests running for 10days before checking
drift, but to do so requires a Caesium frequency reference which -
alas - I lack!
I'm aware that there's a lot of knowledge of these HP oscillators
around here and would be grateful for any pointers.
Alan
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Moin,
On Sun, 31 Mar 2024 14:47:33 -0400
Bob Camp via time-nuts time-nuts@lists.febo.com wrote:
Your OCXO (pretty much any OCXO) goes from room to “pretty hot” ( that might be 20C up to 90C)
in a short period of time. The temperature coefficients of expansion of the various elements of the
crystal never match up perfectly. The same is true for the “stuff” in most of the other components.
This stress translates to a change in value for the components and a change in frequency for the crystal.
To add to Bob's explanation: This also goes for the oven being constantly on
and just the oscillator being power cycled. While the crystal's temperature
does not change (much), it is still a mechanical device that went from being
in storage (albeit hot), to moving. I.e. a lot of equlibrium points just
shifted and need time to relax again. Thus, if you want an OCXO to be
ready to use, while not powering the whole instrument, you still have to
power its oscillator as well.
On Mar 31, 2024, at 10:59 AM, alan bain via time-nuts time-nuts@lists.febo.com wrote:
The manual for the 10811A suggests running for 10days before checking
drift, but to do so requires a Caesium frequency reference which -
alas - I lack!
You don't need a caesium standard to check for drift of an OCXO.
A simple GPSDO will do. All you have to do is measure over several
hours, maybe a day and average the data. If you have a means to measure
the phase (e.g. use the PPS of GPS to start the measurement and the 10MHz
output of the OCXO to stop it), then you are limited by the noise of the
GPSDO only, which averages out pretty quickly and you'll be limited by
the OCXO's instability for anything longer than 1k-10k seconds.
Attila Kinali
--
Science is made up of so many things that appear obvious
after they are explained. -- Pardot Kynes