[time-nuts] Thunderbolt mounting

Magnus Danielson magnus at rubidium.dyndns.org
Thu Jul 12 08:43:50 UTC 2012


On 07/12/2012 05:54 AM, Chris Albertson wrote:
> On Wed, Jul 11, 2012 at 6:55 PM, Mark Spencer<mspencer12345 at yahoo.ca>wrote:
>
>> I tried putting my thunderbolt in a styrofoam box.   It got hotter than I
>> liked and didn't seem to perform any better than when I left it in a
>> cardboard box.
>>
>> I worry about the long term implications for component life as the
>> temperature goes up.
>>
>>
> The advantage of the insulation comes in when you build the fan controller.
>    Mine uses a temperature sensor and an IC comparator that drives a
> transistor that drives a 12V fan.  The fan does on when the set point is
> reached then goes off.  If you set the operating point a little higher then
> the inside of yur house then the fan cycles and keeps the inside of the box
> and roughly, more or less constant.   The parts to build a fan controller
> are about $5

You *really* don't want an on/off regulation scheme. The OCXO gets 
dynamic thermal stress from it which causes it to go off in frequency as 
the heat wave (or cooling wave) comes in and has a rate faster than the 
oven can steer. The oven gain will be much less than for slow 
semi-static temperature changes. Old ovens used on/off schemes, but it 
was dropped as electronics allowed for continuous feedback loops.

You will see a bump in your ADEV at about the cycling period of your 
fan. If you have quick cycling period, it will affect your close in 
phase noise and ADEV, but you might suppress your room/house AC/heating 
loop. If you run continuous, there will be a much quieter response.

Another side-effect of cycling power for an OCXO is that it will for 
each cycle shifts its phase. The OCXO control will return it to 
temperature as fast as it can, which should return it to frequency. What 
happens is that the curve around the balancing point isn't completely 
symmetric, but the heat-up/cool-off temperature profile certainly isn't. 
The end result becomes that the frequency error under the curve isn't 0 
over such a cycle, and that integrated over time will become phase. So, 
this unstable phase creep will keep the control loop active to fight it 
back, until you reach holdover, when the error will be exposed completely.

I've learned this the hard way, from observing frequency and phase 
fluctuations and fighting them.

> If you don't have a temperature controlled fan then the next best thing is
> a well vented cardboard box whose only purpose is to keep air currents off
> the unit. Insulation without an active temperature controller is only going
> to make it hotter, not more stable

Using a cardboard box or other "wall" around the oscillator will as a 
passive oven work well, if the unit is still allowed to dissipate it's 
excess heat to the surrounding. The effect is really great and helps the 
control loop as temperature shifts occurs more gradually such that the 
control loop can track it within it's bandwidth. Long term temperature 
shifts like the bang-bang regulator of the building AC/heating will 
still eat into the box.

Cheers,
Magnus




More information about the Time-nuts_lists.febo.com mailing list