[time-nuts] Z3805A cooling requirements?
magnus at rubidium.dyndns.org
Wed Dec 12 02:54:47 UTC 2012
On 12/12/2012 02:55 AM, Tom Van Baak wrote:
> The telecom closets and data centers I've visited have a significant amount of airflow. Could it be that there is an assumption that these telecom-rack GPSDO expect some level of air?
Well yes and no.
Traditional telecom is supposed to work with self-convection. That may
have worked reasonably well, when they where doing 300W per rack.
At 3 kW per rack, forced air for some units at least is needed, and as
you go up from there, it will be needed for all.
For modern equipment, it's maybe not the power consumption per unit, but
the power density which causes problems.
So, I would expect the unit to assume *some* convection, but not
significant forced convection, and certainly not clogged inputs.
In large datacenters, you need to consider that side-to-side airflows
can cause sequence heat-up as the air go sideways from rack to rack. In
one they had the same switch unit on the top of the rack, and they died
in the direction opposite to their side-by-side air-flow, as hottest air
got killed first.
Anyway, OCXOs doesn't like forced convection, but do enjoy self
convection for best performance.
More information about the time-nuts