[time-nuts] cheap 5V OCXO in 14DIP has about 1E-9 drift per day

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sun Apr 17 19:13:08 UTC 2011


As long as the divisor isnt too large such behaviour doesnt happen.
When the divisor is too large and the filters detune too far then stable 
operation may not be possible.
Until recently the reason for the demonstrated stability of regenerative 
dividers has been poorly understood.
Non linear analysis is required as a linear analysis can lead to 
conclusions that conflict with the observed characteristics of a 
regenerative divider.
To understand the stability requirements one has to examine the 
transient response and the phase portrait of the signals involved:

http://www.its.caltech.edu/~kaushiks/KS_RFIC.pdf 
<http://www.its.caltech.edu/%7Ekaushiks/KS_RFIC.pdf>

http://www.its.caltech.edu/~kaushiks/KS_TCAS.pdf 
<http://www.its.caltech.edu/%7Ekaushiks/KS_TCAS.pdf>

In practice the behaviour of regenerative dividers is sufficiently 
stable and well established that they are being considered for use in 
various atomic frequency standards by NIST and others.

Bruce

Chris Albertson wrote:
> My question about these regenerative filters is that while I know F1 +
> F2 = Fin I'm still wondering how stable it is and how you know your
> divider will not do something like
> 10.0001 + 15.9999 = 26.000  for a few hours and then drift over to
> 9.9999 + 16.0001 = 26.000.    In other words I can see how the filter
> keeps the sum locked to the 26.000 reference but I don't see how it
> keeps the 10Mhz component stable.
>
>
>
> On Sun, Apr 17, 2011 at 3:16 AM, Magnus Danielson
> <magnus at rubidium.dyndns.org>  wrote:
>    
>> On 04/16/2011 10:50 PM, Bruce Griffiths wrote:
>>      
>>> Bruce Griffiths wrote:
>>>        
>>>> Oz-in-DFW wrote:
>>>>          
>>>>> On 4/9/2011 11:29 AM, Greg Broburg wrote:
>>>>>            
>>>>>> <deletia>
>>>>>>
>>>>>> I expect that I am missing something obvious here
>>>>>> a little nudge may help.
>>>>>>
>>>>>> Regards;
>>>>>>
>>>>>> Greg
>>>>>>
>>>>>>              
>>>>> What you are missing is that the concept only applies to small integer
>>>>> (2 or 3) division ratios and won't work as speculated here. It's sort
>>>>> of (long stretch here) like injection locking in reverse. If you want
>>>>> I'll try and post some links to papers later.
>>>>>
>>>>>            
>>>> Nonsense, its already been done for much larger ratios and they need
>>>> not be integers.
>>>> Try simulating it.
>>>>
>>>> Bruce
>>>>
>>>>          
>>> One counter example to the simplistic statement about the operating mode
>>> of a regenerative divider being restricted to division by small integers
>>> only, is that such analysis appears to preclude the possibility of using
>>> a regenerative divider to produce a frequency comb. Unfortunately a
>>> regenerative divider has already been used to produce a low noise
>>> frequency comb where the comb frequency spacing is f/n(where f is the
>>> input frequency and n is an integer). Its possible to extract a
>>> frequency that is a rational fraction (m/n where m and n are integers)
>>> of the input frequency from such a regenerative frequency comb. Thus
>>> there is at least one method of using a regenerative divider to produce
>>> a 10MHz signal from a 26MHz signal.
>>>        
>> As I recall it, in the generalized regenerate divider where two frequencies
>> is filtered these match up
>>
>> http://tf.nist.gov/general/pdf/1800.pdf
>>
>> The two frequencies f1 and f2 has the sum of the input. This has the
>> side-consequence that
>>
>> f1 = fin - f2
>> f2 = fin - f1
>>
>> which is also the conversion steps that the phase will experience over two
>> turns around the loop. For synchronous operation the aggregate phase becomes
>> 0 degrees (modulus 360 degrees).
>>
>> Considering that fin = 26 MHz and f1 = 10 MHz we can conclude that f2 needs
>> to be 16 MHz.
>>
>> As for avoiding asynchronous operations the above NIST articles gives some
>> addtional hints on page 3, among which keeping the loop short is among the
>> important onces, essentially that the electrical delay length doesn't
>> support many modes. Keeping all traces on a normal PCB for 10 MHz and 26 MHz
>> should avoid that issue completely.
>>
>> This would form a 5f/13 - 8f/13 system since 2 MHz is the common frequency
>> for all of these. Keeping phase solutions unique for 2 MHz separation should
>> not be too hard.
>>
>> Cheers,
>> Magnus
>>
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to
>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>>
>>      
>
>
>    




More information about the time-nuts mailing list