[time-nuts] Achievable temperature stability for Thunderbolt

WarrenS warrensjmail-one at yahoo.com
Tue Jan 18 11:07:14 UTC 2011

Achim, DH2VA

AFTER temperature is being controlled pretty good with the default settings 
(modified if necessary),
What I use to fine tune the PID values is the "Quick Auto Tune" command 
The quick command is much faster (typ < 30 min) than the full Auto Tune 
command, and can and should be used even with a hack and whenever there is a 
change in the Temperature H/W.
(The "quick auto tune" command skips a lot of steps by assuming that the 
temperature is at the set-point and under reasonable stable control and not 

To set the PID values to the slow default use "K,W,Enter", for the faster 
default settings "K,M,Enter"  (there may be others?)
For best results, I keep the inner case that holds the sensor & OCXO closed, 
and do not let air blow inside it.
Note, unlike passive temperature control where the more the mass the better, 
that is Not the case for this simple active PID temperature controller.
The more the mass, the slower the response, and the harder it can be for the 
temperature loop to stabilizing.



[time-nuts] Achievable temperature stability for Thunderbolt environment?
Achim Vollhardt avollhar at physik.uzh.ch
Tue Jan 18 08:21:48 UTC 2011

Dear Warren and Mark,

I have hacked together a very simple setup with the transistor circuit
and the opened Thunderbolt case (not its own case, but the front lid of
my outer case opened). The CPU fan was placed about 2" away, only
blowing indirectly into the case..

The results are simply amazing! The temperature pk-pk excursions are
down to 0.03K, approx. 30x less than before.. and I even have not
figured out how to tune the controller parameters..
.. which will have to wait. I don't think is useful to tune a hacked
setup. I will add as much thermal mass (aluminum) into the case as I can
and do a proper fan installation.. then back to PID tuning.

Thank you very much for this great tool! I wish I knew C/C++ better so I
would have less problems understanding the code..

73s Achim, DH2VA 

More information about the time-nuts mailing list