[time-nuts] Re: Envornmental control
phk at phk.freebsd.dk
Wed Mar 9 17:43:22 EST 2005
In message <006901c524f6$6d6c5c60$4715f204 at computer>, "Tom Van Baak" writes:
>I'm cc'ing the list since I'd be interested in other
>stories about temperature control.
I visited the Danish Fundamental Metrology lab some time ago, and
had a chat with them about a lot of stuff, including temperature
control (they need it for calibration of standards resistors
and all sorts of other stuff).
It is important to realize that it not enough to control the
temperature of the air, you also need to control the flow and finally
the humidity is very important because it seriously affects the
airs ability to absorb heat.
They also gave me the following cheap hint about environmental
control: Find and old fridge and use it as an insulated chamber.
(The compressor doesn't have to work: it is not used).
Provided the equipment you want to have constant temperature does
not produce too much heat (ie: OCXOs etc), putting it 1/3 from
the bottom of an old fridge with a slow moving fan 1/3 from the
top to keep the air moving (slowly!) and putting the fridge
in a room with reasonable thermal stability, you should be able
to get milli-kelvin temperature control.
For equipment which generates more heat it is harder because
you will need active temperature control, and the task becomes
close to impossible as the heating goes up.
For moderate heat output (a Cs ?) the fridge method can still
be used but with a twist: Take the compressor out and fill the
circuit with glycerol (water would make it corrode) put a small
pump on it (aquarium stuff) and use fans/peltier elements to control
the temperature at the inlet to the fridge. (I'm wondering if not
some of all the "overclocker watercooling" hardware out there would
Some experimentation is necessary to find which fluid temperature
gives the desired inside temperture. Again, remember to have a
slow moving fan on the inside to keep the air moving in a steady
If the power dissipation of the equipment is (close to) constant,
like for a HP5061, then nothing should be needed. If the power
dissipation varies, it may be necessary with a temperature sensor
and active feedback.
I havn't gotten around to try this myself, but I would probably put
a couple of Dallas 1-wire temperature sensors and maybe a webcam
in there as well, so I wouldn't have to open the door to check the
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk at 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.
More information about the time-nuts