[time-nuts] Morion MV89A position
Charles Steinmetz
csteinmetz at yandex.com
Wed Jan 29 20:28:32 UTC 2014
/tvb wrote:
>I tried something like that and ended up frying the oscillator! The
>problem was that I used too much insulation. That was good to keep
>ambient temperature transients out but it also kept internal heat
>generation in.
> * * *
>My question is, how does one design an enclosure to prevent this
>mistake? Or is it trial and error. Perhaps put a thermistor on the
>OCXO and if the case temperature rises beyond what is normal case
>temperature in free air, then the enclosure has too much insulation?
That is the point of distinguishing thermal capacitance (= thermal
mass) from thermal resistance. Generally, OCXOs are thermally
designed to be "naked" in some abient environment that is close to
room temperature (usually somewhat above, anticipating they will be
installed in warm instruments and not be sitting naked on a
desktop). They will work acceptably in an ambient environment
ranging some tens of degrees C above and below this, as long as the
ambient temperature does not change too fast for the oven controller
to keep up (for example, the HP 10811 is rated for -55C to 71C, with
degraded temperature stability below 0C).
Adding additional thermal resistance (your insulation, or Paul's
Dewar) unbalances the heat flow to ambient. In theory, the
oscillator should respond by backing down the heater current to match
heater power to the new thermal resistance. However, raising the
thermal resistance can destabilize the thermal control loops in some
OCXOs. When that happens, you get poor thermal regulation at best
and you burn down the oscillator at worst. (And even if the control
loop does not destabilize, it won't be running at the design center
"sweet spot" foreseen by the designers, so thermal stability will
likely be degraded.)
What you want to do is keep the net thermal *resistance* similar to
what it is with the oscillator in its target environment, and add
thermal *capacitance* to slow down the changes in the oscillator's
immediate environment (i.e., just outside the oscillator can). The
metal box does just that. The thermally resistive path is
essentially still all air, just as it is with the oscillator sitting
on a desktop -- from the oscillator through the air surrounding it to
the inside wall of the cast box, then from the outside wall of the
cast box through the air to ambient (to a first approximation, the
temperature of the inside wall of the box is the same as the
temperature of the outside wall). This way, the average thermal
resistance from the OCXO to ambient is still similar to what the
thermal designers contemplated. However, the cast box averages
(integrates) the outside temperature as seen by the OCXO with a time
constant measured in tens of minutes.
The net result is that the oscillator is in (or very close to) its
design environment with respect to the average heat loss, so the oven
controller is working at (or very near) its design sweet spot -- but
at the same time, the rate of temperature change seen by the OCXO
(i.e., its immediate environment inside the cast box) is integrated
over tens of minutes. If the ambient temperature did not change at
all, the OCXO's immediate environment inside the box would be exactly
the same as if it were not in the box but, rather, sitting naked on a
desktop. When the ambient temperature changes fast, the OCXO's
immediate environment follows it -- but much more slowly, giving the
oven control loop time to adjust to the change without introducing a
transient error.
As long as you end up with sufficient thermal capacitance to slow all
expected thermal transients down to the point that the oven control
loop has no trouble keeping the quartz crystal in an isothermal
state, you have done all you need to do. There is nothing to
calculate, and no complicated design procedure. If the oscillator
still exhibits some transient thermal error, just increase the mass
of the box by mounting aluminum plates to the outside walls of the
cast box until it doesn't.
The one exception to this would be if the oven controller simply has
too little gain to keep the crystal isothermal, in which case there
would be a DC (not transient) error (each ambient temperature would
be associated with a unique crystal temperature). Note that this
would not be the fault of the method -- rather, it would be a matter
of poor regulation by the OCXO's oven controller. It could be
mitigated by active regulation of the cast box temperature using an
outer enclosure with a thermostatically controlled fan.
Best regards,
Charles
More information about the Time-nuts_lists.febo.com
mailing list