[time-nuts] Thunderbolt oven / non-stable operating temperature

Charles P. Steinmetz charles_steinmetz at lavabit.com
Tue Dec 11 04:23:13 UTC 2012


>That temperature sensor does have an effect on the final outcome as 
>it is part of
>the internal equation.  So buffering the ambient temperature is important.

I've heard this before, but the evidence I have seen does not seem to 
support the proposition.

While switching the Dallas chip in one, I used the opportunity to 
bring the chip temporarily outside of the Tbolt housing on a cable to 
investigate whether the Tbolt makes any internal use of the 
temperature data.  Neither freeze spray nor bringing a soldering iron 
near the chip, when it was outside of the Tbolt housing and the Tbolt 
housing was well insulated from the changes in chip temperature, 
seemed to have any effect on the operation of the Tbolt, either 
normal or in holdover.

I have also run Tbolts with the newer ("wrong") temperature chips for 
long periods, and have not observed any systematic differences in 
performance between them and units with the older chips, either in 
normal operation or in holdover.  In Tbolts with the newer chips, the 
reported temperature often has little connection with the actual 
temperature and, at times, jumps abruptly, yet the Thunderbolts 
operate normally with no corresponding jumps in operating parameters.

My supposition/conclusion is that the temperature sensor was provided 
so telcom operators could get a rough idea of the temperature in 
remote cell-site transmitter shacks, not for internal use by the Tbolt.

As long as the Tbolt is housed so that its reported temperature does 
not change too rapidly, the oven control loop will keep the crystal 
very close to its set temperature over a wide range of ambient 
temperatures.  I have used this approach and have also actively 
controlled the housing temperature, and have not observed any 
material difference in frequency or timing stability between the two 
approaches.

Best regards,

Charles











More information about the time-nuts mailing list