[time-nuts] Airflow around oscillators.

Bill Hawkins bill at iaxs.net
Sun May 22 13:08:49 EDT 2005

Well, it's generally not a good idea to put the heat producers
in the same box with temperature-sensitive stuff. Have you
calculated the heat dissipation from components (or measured
it from the mains) and balanced it with the thermal resistance
of the box? What temperature rise do you get?

If you do not insulate the box from ambient temperature swings
(also not a good idea below ~10E-7) then the extra heat rise
caused by the power components will reduce the size of the
swings if the box gets to 50 C or so. Your batteries will not
be happy.

It's really not a good idea to put batteries in the same box.
Yes, they're sealed now, but one day one of the cells will be
shorted by debris, at the end of its 5-10 year life. Your
charger will put max current through the remaining cells as
it attempts to hold 13.7 volts. The battery case will expand
from gas pressure and finally crack one or more vents. But
perhaps you don't expect your current fascination with precision
time to last that long.

If you put thermal insulation around the oscillators then the fan
will be of little use. Poul-Henning Kamp's suggestion of a 
relatively large insulated box with some internal air circulation
is excellent. If you do that, use three boxes - one for power
supply, one for oscillators (spread out so the ovens don't interact)
and one for the batteries. It would be best to have a charger that
senses battery temperature. Bring the charging current out to an 
external meter so that you can monitor it.

Remember that PID is feedback control. An error must exist before
the output will move. Tuning for precise setpoint control is not
always best for controlling measured disturbances. It is best that
the disturbances have slow rise times to avoid driving the PID
into an unstable region. Now, if you could mathematically model
the effect of ambient temperature, you could measure ambient and
pass it through the time-dependent model to become feed-forward
for the oscillator ovens. You'd have to rebuild the oven controllers.

Bill Hawkins

More information about the time-nuts mailing list