[time-nuts] temperature sensor

Charles Steinmetz csteinmetz at yandex.com
Mon Jul 21 18:23:31 EDT 2014

Tom wrote:

>There have been several discussions over the years about variable 
>fan speed based temperature control. I can't explain it, but I've 
>always been suspicious of this technique. It seems to me still air 
>is inherently better than moving air. Passive (no fan) is better 
>than active (fan). And constant velocity is better than turbulence 
>is better than variable velocity. But I don't know for sure.

There is no such thing as "still air," unless there is no temperature 
gradient.  If there is any temperature gradient (typically due to 
power dissipation), there will be convection currents.  In a closed 
space (for example, internal to a TBolt or in a sealed box that 
encloses a TBolt), these convection currents will set up a flow 
pattern that may be benign or malicious with respect to keeping a 
particular part of the device at a constant temperature.  If the part 
you are particularly interested in is a creator of thermal gradients 
(as the OCXO in a TBolt is), analyzing this gets very complicated very fast.

Fans (speaking here of fans that circulate air internal to a closed 
volume, not fans that exchange air between the inside and outside of 
a volume) tend to mix up the air and reduce thermal gradients.  Then, 
the question becomes whether the circulation due to the fan has a 
patterned or a random thermal flow.  Typically, a random (diffused) 
pattern is best -- but it is relatively hard to achieve.  With 
careful design, active circulation is usually better than passive 
convection.  However, "careful design" is not easy.  Also, fans raise 
a concern about vibration, which is a real worry with any precision oscillator.

One other possibility is to use passive techniques to randomize (more 
or less) the passive convection.  This can be achieved (to a degree) 
by filling the internal volume with low-density, very porous 
insulation.  On a larger scale, a sealed box of, say, 2 cubic feet 
can be filled with common packing peanuts and the isolated object 
placed in the middle.  Air will still circulate by convection, but in 
a more random manner.  (There will also be less bulk circulation, so 
the thermal gradient will be somewhat larger than before.)  Applied 
to a TBolt, one might fill the inside of the TBolt itself with 
smaller pieces of styrofoam (irregular shapes perhaps 6 or 7mm in 
size).  [Spheres (styrofoam beads) may pack a bit too tightly for 
this, impeding airflow more than desired.]  The same can be done for 
a sealed box that encloses a TBolt or other oscillator.  I have 
achieved very good results with this method, when properly applied.

I have done a fair amount of experimenting with and without fans (but 
one must recognize that there are so many variables, even a lot of 
experimenting really only scratches the surface), and have always 
found that passive circulation (within sealed volumes) works very 
well when the object ultimately being controlled is an ovenized 
oscillator.  For tight control, which is needed for precision voltage 
references, DAQ circuits, and other precision process-control 
applications, I do use a thermostatically operated fan to exchange 
air between the outermost sealed volume and ambient -- but even this 
I usually find unnecessary if the ultimate object is minimizing the 
frequency drift of an ovenized oscillator.

Finally, re.: fan control.  For a brushless DC fan to run slowly, you 
need to feed it full voltage with pulse-width modulation 
("PWM").  Even then, they will not run all that slowly.  The 
Microchip TC642B fan controller (8 pin IC, about $1.20) is a very 
handy part when you need a wide range of fan speeds.  It uses 
commutation noise to sense fan rotation, and has a "stall routine" 
that gives the fan a kick if it stalls (NB: this is a feature of the 
642B, absent on the 642).  So, not only will it run the fan at its 
lowest possible self-sustaining speed, you can also run the fan much 
slower than its self-sustaining speed by letting it stall and be 
restarted periodically.  The fan looks like a windmill with three 
sheets to the wind below its self-sustaining speed, but it works 
extremely well and this operation does not damage the fan or the controller.

Best regards,


More information about the time-nuts mailing list