[time-nuts] Strange oscillations

Bill Hawkins bill at iaxs.net
Sat Sep 4 11:07:14 EDT 2004


Alberto di Bene wrote:
"Which can well be in the hysteresis band of the oven proportional
temperature controller."

Been in Industrial Process Control for 40 years, so here are a few
observations. They may be completely wrong because I can't test the
system.

Proportional control does not have any hysteresis. The thermistor
and the oven control should also be free from hysteresis. You are
correct that you need hysteresis to have a limit cycle. What you
have is a limit cycle because the amplitude is relatively stable.

Here are some other possibilities:

Your reference oscillator is cycling. What are you using to measure
to 10E-11 accuracy?

Your line voltage (heater supply) is cycling. Proportional control
needs some error to change the heat to the oven, so it is sensitive
to changes in all ambient conditions. Can you measure the heater
supply voltage to 10E-11? Probably not, but if the proportional gain
is 100 you only need 10E-9, still an expensive measurement. Try using
a battery for the oven supply.

"The OCXO is not yet closed in the box that will finally contain the
project, it is still exposed to the ambient air."

It is possible that the designers were counting on the enclosure to
stabilize the oven. If the controller is actually proportional plus
integral (and maybe plus derivative) then the controller tuning may
be too sensitive for an oven exposed to ambient air. Try insulating
the OXCO with packaging foam or fiberglass so that it is completely
enclosed except the wires. That is easier than adding thermal mass,
but if the OXCO is supposed to mount on a plate, mount it on a plate
about twice the area of the mounting surface. Copper would be nice
but aluminum is cheaper.

Remember that you are making a very sensitive measurement. Your
presence in the room could change the heat required because you
radiate about 500 watts in the infrared.

Please let us know the results of your experiments.

Regards,
Bill Hawkins





More information about the time-nuts mailing list