[time-nuts] Effects of filter delay

Dr Bruce Griffiths bruce.griffiths at xtra.co.nz
Sun Dec 24 18:03:51 EST 2006


SAIDJACK at aol.com wrote:
> Hi Bruce,
>  
> yes the heater circuit is problematic, and will affect stability a  lot.
>  
> For example, some OCXO's don't have separate (Kelvin sensed) ground pins  for 
> the heater, and EFC voltage.
>  
> The result is that any changes in the heater current will cause a voltage  
> change in the ground pin of the OCXO from the OCXO body to the PCB ground plane  
> (due to the finite resistance of the ground pin).
>  
> This I2R loss is very small, but will superimpose onto the EFC voltage, so  
> it's as if this voltage is placed onto the DAC output.
>
> We have measured this change of voltage to be in the order of 10's of  
> microvolts for some OCXO's, and that translated into 10's of EFC DAC steps  worst 
> case (during power-on for example when the heaters are running at maximum  
> current).
>  
> Again the problem is not the steady-state heater current, but changes in  
> heater current due to thermal tracking, and loop stability of the heaters, this  
> in turn causing changes in EFC voltage. We are talking about 1E-012 or  better 
> being desirable stability, so these marginal effects almost sudden become  a 
> real issue.
>  
> The best approach is to separate the ground return current of the heater  
> circuitry completely from the DAC ground circuitry. But that's not possible on  
> many OCXO's...
>  
> Bye,
> Said 
> _______________________________________________
> time-nuts mailing list
> time-nuts at febo.com
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>
>   
Said

I guess this means you have to model the effect of the changes in IR 
drop due to heater current changes and modify the EFC voltage to compensate.
The complicating issue being that the value of R will vary with ambient 
temperature (copper has a resistance tempco of about 0.4%/K).
Since the a compensation accuracy of around 1% or so will suffice, 
analog compensation circuitry could be used, however it would have to be 
adjusted to suit individual oscillators.

Bruce



More information about the time-nuts mailing list