Hi

People on the list have written big long papers about just why ovens do what they do.
To get into all the details, those papers are a much better reference than some quick
messages on an forum like this.

The quick and dirty version:

There are a number of things that impact the temperature stability of an OCXO. Gradients
are a big factor. Your heat source, sensor and resonator are never all at a single microscopic
point within the device. Tear one open and that’s what you will find. Heat comes from over
here, sensor is on that side, blank in the crystal package is here and has a thermal path to
this over here.

If you take an SC and fiddle the trap so you are in “thermometer mode”, you can measure the
thermal gain to the crystal. You will find that as you do this or that the gain goes up. Eventually
it changes sign and comes back down again. The higher it gets, the more likely some very minor
thing will have an impact on the gain ( = it’s a sharp peak ).

From talking to the folks who did the Z3801, they did not have the knowledge or skills to dig into
the insides of the 10811. They also did not have a need to do so. Their task was simple. All
that was required was to put a “blanket” over the device to make it warm up from a colder
temperature. Once that warmup was done, their heater stoped doing anything useful.

Since it was not designed to improve the stability of the OCXO, leaving it turned off will
not degrade the stability.

Bob

Hi

Other than sitting down and talking to the folks who actually designed it and who
ran the design process, no I don’t have any direct info. I do know that, based on 40+
years as an OCXO designer that the approach used is nonsense. You could do a much
better job if stability was the target. Since it was not why they did it this way, that’s in
no way a knock on those folks. They did what they needed to do.

Bob

Hi

The 4C rise with the box in place is a measure using a real box. Simply put, the stuff inside
runs at about 4C higher with the box in place relative to the room temperature than it does
with the box missing. That’s after letting things sit for many hours …. (and checking after
a few days). Given all the variables, yes, it could be 5 or 6C. It’s certainly not anything less
than 4C. The room cycles so picking a target is a bit wonky.

The box deliberately has a somewhat open top, to keep some level of convection cooling
what’s inside. Yes, that’s another un-mentioned detail and it certainly has an impact. Lots
of details like that would get involved in a precise answer to the question. One could get
all sorts of cool modeling software involved.

Right now, all I’m really looking for is an order of magnitude sort of answer. Put it in terms
of whatever your favorite material is:

How many liters or kg of this or that will it take to do the job of knocking the swing down to
under 1C over 24 hours?

That’s with the room swinging 2 to 4C over the same period.

The limit on insulation is pretty obvious: Put to much on and the device(s) inside the box
overheat and may be damaged. Some significant margin between max ambient on the
device(s) and box temp needs to be maintained.

The limitation on the thermal mass really is however much room you happen to have.
Heading down the mass route without a rough idea of what’s going to be needed does
not sound like a lot of fun. In my case a redesign of several things comes in each time
you adjust dimensions.

Obvious alternatives get you right back to some sort of servo setup running this or that
as a control element. (unless there is something else ….).

Most of the “big guys” seem to have found a convenient cave (or built one). Then then
get to publish papers on the shortcomings of thermal control in their custom built
cavern ….. All of the things they complain about apply here. The power load in the
area is variable. People (and dogs) wander through on a random basis. There’s a door
to an unheated space here and another one over there. They get opened from time
to time …. (none of that goes into the 4C number, but probably should ….. )

Bob

ed breya via time-nuts writes:

Unless you have cranked your fan or A/C up to gale force winds, you can safely ignore that inside a building,

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Watching this discussion of laboratory temperature control, I feel that my experience might be helpful. I inherited a laboratory, about 1980, about 4 X 5 X 7 metres in dimension. I had the help of a refrigeration/AC tradesman, and together we wanted to try to improve conditions.

There was a ~ 1 HP reverse cycle window conditioner in the middle of the long wall.
Measurements with a couple of thermohydrographs (clockwork recording temperature/humidity recorder)
showed swings of 6 or 8 degrees C in the course of a few hours. The temperature was controlled by a  mechanical switch which had hysteresis of about 2 to 4 Degrees C. The unit operated like this: for about an hour the compressor laboured away cooling the room until the return air to the conditioner
had fallen by 2 or 3 degrees.
Then the unit would switch itself off for a few minutes (to decompress) and then it would chug away for another quarter hour it would heating the air until the room had heated 2 or 4 degrees.
More of an oscillator than a controller.

After some experiments this is what we finished up with.
A small computer fan was fixed to the wall blowing air parallel to the wall. This caused the air to slowly rotate around the room. Air velocity was less that 0.5m/s, which is the threshold to be able to sense air flow against the face. So there was no apparent airflow unless you looked for it.
The air took about a minute to travel around the room. The stirring was enough to make the air behave like a well stirred vessel. Now the room contained about 140 Kg of air, so the the heating by one kilowatt was about 0.007 degrees/second.
The air temperature was measured with a tiny glass encapsulated thermistor which was near enough to the fan to responds in one or two seconds. It turned out to quite accurately measure the average temperature of the room.
The output of the thermistor circuit was simply over or under the setpoint.
The compressor was switched with zero voltage switching industrial solid state relays.
What made this system very successful was the logic connecting the temperature sensor to the compressor, and heat/cool relay.

There were several rules to implement.
If the room was in heating mode;
if the temperature rose 0.1 degree above setpoint, the
compressor was stoped. If the compressor is ever stopped it must stay stopped for one minute.
At the end of the minute, if temperature is below the setpoint, the compressor is switched on.
For the next 5 minutes if the temperature falls below setpoint, the compressor starts. If after only 2 seconds the temperature rises above the setpoint the compressor is switched off. This is not ”short cycling” because the compressor stays off for for a minute.
Cool cycle;
If after 5 minutes of no heating demand the cooling mode is implemented.
The window unit is put into refrigeration mode, the compressor is started and keeps running until the temperature falls below the setpoint. When that is achieved, the compressor cuts out and back in, always allowing a minute(or two) for the gas pressure in the compressor to subside so it does not have to start under load. If no cooling is required for 5 or ten minutes the mode is changed.

The temperature control was miraculous. The temperature did not deviate by more that a small fraction of a degree for weeks on end.
The changeover period, from heat to cool and back, did not cause noticeable transients because they only occurred when conditions outside had cooled or heated to nearly match the setpoint of the room.

5 or 6 of these rooms were built, a paper was published, but it was attacked with papers saying it could not work by people who never tried it, and would not even come to witness it.

So stay away from 3 term controllers, they are not relevant, and commercial “Thermostats”
and use this simple on off controller either ramping up or down at 0.5 degree per minute
and with one second response time trimming swings to a few hundredths of a degree Celsius.

cheers, Neville Michie