[time-nuts] Thunderbolt stability and ambient temperature

J. Forster jfor at quik.com
Tue Jun 9 19:26:34 EDT 2009


>> The point for timenuts is not just the thermal resistance, but more
>> importantly the thermal *impedance*: you want to low-pass filter
>> the thermal changes so that they all happen in the area where the
>> PLL can cope with them.
>>
>> Thermal resistance is about insulation, thermal impedance is
>> about (thermal) mass.

Not really.

Consider a one dimensional model, with the controlled space onb the left
and the ambient on the right:

     |------|           |
     | Osc  |===========| Ambient
     |======|           |
      T mass

> The thermal time constant (not the thermal impedance per se) is what
> matters when one is trying to reduce the effective amplitude of
> temperature fluctuations due to air conditioner cycling.
> Adding mass increases the thermal capacitance adding insulation
> increases the thermal resistance.

True

> It is possible to construct an enclosure with a long thermal time
> constant together with relatively low thermal resistance so that the
> temperature of a GPSDO or similar device within the enclosure only
> increases by a relatively small amount.

Nope. This is essentially a thermal low pass filter. The same resistance
enters into the time constant and thermal resistance to the ambient. The
only way to increase the thermal TC, while maintaining the resistance, is
to add thermal mass to the Oscillator assembly. (Increase the C with
constant R)

-John


> Multiple alternating layers of thermal conductor and thermal insulator
> reduce thermal gradients as well as temperature fluctuations.
>
> Having an outer conductive layer reduces the temperature gradients over
> the insulator surface.
>
> Readily available inexpensive aluminium foil is a cheaper alternative to
> expensive noble metal foils.
>
> Silica aerogel is one of the most effective insulators.
>
> Balsa wood has been used as the insulator in portable temperature
> controlled ensclosures for standard cells.
>
> Bruce






More information about the time-nuts mailing list