My understanding of thermal regulation is that one of the more difficult thing to do is precisely to know what is the temperature of the components you are trying to stabilize.
If you have one oven (even double oven), you have one temperature sensor. The best you can do is keep the sensor at a constant temperature. Assuming the temperature sensor is kept at a perfectly stable temperature, how constant will the temperature of other components be will be a matter of hardware design and layout. I.e. the components you are trying to stabilize should be closely thermally coupled to the sensor. Because there are more than one component to stabilize, not just the crystal, it is a difficult task. The issue is not just static regulation, but also regulation in the presence of a temperature gradient. We understand that if you change the ambient temperature by X, the oven will react to keep the sensor's temperature constant, but the distribution of that heat flux will change dynamically (particularly it will be affected by the masses that are in the way, between the source of heat and the components that are to be stabilized), and even if you have perfect static regulation, in most cases there will be a non-zero dynamic response (a transient). The designers of the HP E1938 (which never went to full production) went through pains to try and keep the gradient evenly distributed precisely for that reason. My guess is that it you take the cover out from the E1938, you will find a perfectly symmetrical layout around the center, where the thermistor is located. A large massive cover over the OCXO will go a long way to reduce the transient and allow the oven controller loop to react smoothly, but as long as the time constant is less than the period of variation (a day?), the peak to peak frequency variations will not change much, but they should be more predictable and may be compensated by a feed-forward mechanism, or a Kalman filter. Didier > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Magnus Danielson > Sent: Saturday, July 12, 2008 7:50 AM > To: [email protected] > Subject: Re: [time-nuts] Double ovened 10811-60158 on ebay > > From: "Ulrich Bangert" <[EMAIL PROTECTED]> > Subject: Re: [time-nuts] Double ovened 10811-60158 on ebay > Date: Sat, 12 Jul 2008 13:09:55 +0200 > Message-ID: <[EMAIL PROTECTED]> > > Ulrich, > > > > I keep wondering if not a passive oven (metal box, > insulation, metal > > > box) would be sufficient. Worst case temperature change > rates would > > > be significantly reduced such that the oven loop can track it > > > better. The remaining temperature shift will be less. > Basically acts > > > like a lowpass filter. > > > > I have been thinking pretty much the same and therefore I > built me an > > "outer" box for my 10811 that has abt. 2-3 cm of free air > between the > > outer surface of the 10811 and the inner surface of the box to make > > the 10811's temperature regulation "look" into the same "thermal > > leakage impedance" that the designers may have in mind when they > > decided for the controller's parameters. > > > > Then comes 2 cm of massive aluminium which's heat capacity - in > > conjunction with the surrounding air's thermal resistance - is > > expected to give the desired "lowpass filter" effect. > > > > Due to the nature of my disciplining system I have been able to > > measure the overall effect of this shielding quite precise: The > > lowpass effect is well defined by > > > > a) an increase of "phase delay" between temperature changes outside > > and changes of the oscillator's frequency that is now in > the order of > > some hours. > > Which does not as a big supprice but rather confirms the expected. > > > b) the frequency changes (as far as they seem to be related to > > temperature) appear to be "smoother" and more predictable > > Which also matches expected and other measurements done. > > > Nevertheless I have been a bit disappointed that the > overall frequency > > changes along the diurnal changes of the surrounding > temperature in my > > flat have been pretty much the same in terms of amplitude with the > > above mentioned phase delay. > > Your time-constant for the lowpass filter does not do much to > the diurnal frequency. The temperature changes still go > through. The point with the passive oven is that it does > smooth things out, and could potentially make the oven design > inside it an easier task. If temperature changes is the main > problem then it may be all that is needed to get the extra > margin. It is certainly not a perfect solution, but an > interesting exercise which may be of help for some cases. > > > I have then started to compute what thermal time constant may be > > expected from this arrangement. This is not trivial stuff > because in > > contrast to the aluminium's thermal properties their > counterparts for > > the surrounding air are not well defined. As far as I > remember, a BEST > > case assumption gave a result of 2-3 hours that seems to match the > > empirical results. If you are out for time constants that > are able to > > smooth diurnal changes I guess you will have to throw in MUCH more > > material. > > You would also like to build multiple stages to form a higher > degree filter. > The outermost box is really there to create the chamber > around the isolating air and in there your heavy mass (big > cap) is gets heated (charged) and chilled > (discharged) through the air and black-body radiation. > > Having the heavy mass fairly directly exposed to the > surrounding air will provide better conductivity through drag... > > Oh, humidity will change the heat conductivity of the air noticeably. > > Cheers, > Magnus > > _______________________________________________ > time-nuts mailing list -- [email protected] To unsubscribe, > go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
