[time-nuts] Reference oscillator requirements when measuring ADEV ?

Bruce Griffiths bruce.griffiths at xtra.co.nz
Fri Jul 15 00:44:11 UTC 2011


Mark Spencer wrote:
>
> I'm curious in knowing if there are any rules of thumb as to how much better a
> reference oscillator needs to be than the device under test when measuring ADEV
> ?
>   
> To provide a bit of context to my question I'm using a HP 5370B to repeatedly
> measure the time interval between the 10 Mhz signals from a FTS 1050 and
> rubidium oscillator.  
>
>   
> Having compared the FTS 1050 to another OCXO in the past I'm confident that it
> is meeting or exceeding it's spec of an ADEV of 1X10-12 or better at tau's of
> say 100 thru 1000 seconds.   I'm also confident that at tau's of 100 seconds or
> more the noise introduced by the 5370B is less than 1x10-12.   
>   
> The underlying question that I have is how much confidence could I put in
> measurements that show the rubidium of having an ADEV of say less than 2.0X10-12
> at a tau of 100 sec ?  
>
>
> This is just for a hobby (: and I'm not looking for any degree of certainty but
> I'm curious in understanding how accurate the measurements might be and if
> changes in the measurements might not be caused by changes in the device under
> test.  
>   
> Thanks in advance for any replies.
>   
> Regards
> Mark Spencer
>
>
>    
The reference oscillator needs to be at least a factor of 3 (preferably 
10x) better for all Tau of interest.
Unless of curse one uses a 3 (or N) cornered hat technique in which case 
the ADEV of the oscillators being compared may be similar for the Tau 
range of interest.
However one needs to guard against correlations between oscillators due 
to injection locking, environmental fluctuations etc.

For Tau= 100s relatively simple low cost three cornered hat setups (with 
an instrumentation ADEV noise floor below 1E-12) are feasible (provided 
on has a suitable offset oscillator) using inexpensive digital logic 
plus a suitable microprocessor.
No complex interpolators are required.

Bruce



More information about the time-nuts mailing list