[time-nuts] Tbolt temperature sensor

SAIDJACK at aol.com SAIDJACK at aol.com
Thu Feb 5 19:54:10 UTC 2009

Hello Tom,
>The reason I'm asking this is because I can't figure out  from
>your descriptions or from the plots a clear way to  distinguish
>between unintended oven current-induced changes in  applied
>EFC voltage vs. direct temperature-induced changes in  OCXO
>frequency output. It seems both would have the same  effect.

As long as all effects are fairly linear over a given temperature range,  
then it doesn't matter where the thermal sensitivity is coming from. You may  
even have canceling effects (if for example the oven has a negative tempco, and  
the DAC/Vref have a positive tempco). So in other words, even the unintended  
temp-induced EFC changes will be compensated, because we measure absolute  
current versus absolute EFC voltage when the unit is locked, and calculate the  
parameters from this measurement, which includes all unintended thermal  

>I wonder if there is a clear test that would tell you one way  or
>the other.

You could test specifically for the unintended effects: turn off  temperature 
compensation and GPS locking, and put the unit into a thermal  chamber and 
see how it performs. On our double ovens we usually achieve better  to 
significantly overall performance than the OCXO thermal spec itself. This  shows that 
we don't "make things worse" by the DAC, DAC-reference, or grounding  thermal 
Since our PCB can perform better than the double ovens we use,  this means 
the PCB is about 50x to 100x or more better than the  performance of the single 
ovens, since that's the relationship of double/single  oven performance.
One caveat: since the double ovens are so extremely good, it can actually  
happen that the electronic compensation could under some circumstances make  
things worse than if it was just turned-off. This can happen if you operate your  
oven in a small thermal range (say +/-5C) where the current/temp relationship 
is  nice and linear, and then almost sudden go -50C lower than usual, where 
the  relationship may be parabolic etc.
For applications that require large thermal ranges (say fighter aircraft  
that can descend 30000 feet in mere minutes or less) we actually disable the  
tempco statistics, and rather measure the unit over a 100C range in a thermal  
chamber, and establish the best tempco parameter that will work from -25C to  
+75, and will improve the units performance over the entire temperature range  
without adding additional error anywhere.
In all of our Fury and FireFly GPSDO's the user has the option to put a  
hard-coded tempco value, and disable statistics gathering.
To check for ground loops: solder a wide (say 1/2 inch) copper string from  
the OCXO case to the power supplies' ground pin so the oven ground current will 
 flow through this instead of the OCXO ground pin. We do not see any change 
in  performance doing this, and with a current clamp you can verify that most 
of the  current is actually flowing through this new ground.

>Maybe change oven voltage quite suddenly. That should  cause
>a change in oven current. If the measured EFC voltage or  the
>measured frequency also changes suddenly then that  might
>indicate it's not really temperature that's being compensated  for.

This would not work, since it would change the oven current for sure, and  
this would end up causing an immediate EFC voltage change since it would be  
interpreted as a thermal change (which it is not over the long term of course  
after the oven settles down). That's why we have a highly accurate, low noise  
10.45V regulator on our boards to generate the OCXO voltage, and don't use the  
raw supply voltage.

>Can I do these tests on my double oven Fury? Or is this  something
>only for the single oven version.
You can do these tests on the double oven Fury. The tempco numbers are  
usually around 10 - 50 for a good double oven, and around 1500 to 4000 for a  good 
single oven. This means the compensation is usually 100 to ~300 times more  
aggressive on a single than on a double oven unit. You may want to upgrade  the 
firmware to the last release (rev 1.21) for these types of tests.
On a typical double oven we thus see very little correlation between OCXO  
current and EFC voltage locked or unlocked :)

More information about the time-nuts mailing list