[time-nuts] Measuring 10 MHz accurately: correction

[Bruce Griffiths]

Bruce Griffiths wrote:
> Martyn Smith wrote:
>> Hi,
>>
>> Two questions for all of you today.
>>
>> 1st Question
>>
>> I can measure 10 MHz frequency to an accuracy of about 2 parts in 
>> 10E-12 in a one second gate time.
>>
>> I use a SR620 time interval counter and make a timeAB measurement.
>>
>> I make two measurements of phase (between my UUT and my reference)  
>> and then calculate the frequency.
>>
>> I have some nice software that controls the SR620 and does all the math.
>>
>> I'm sure most of you understand my measurement technique, since it's 
>> as old as the hills.
>>
>> Does anyone have a good application note explain this standard 
>> procedure? I'm trying to explain it to a friend, and can't find a 
>> nice application note with some diagrams, formulas etc.  I know HP 
>> did one and Standard did as well, but can't find any quickly.
>>
>> 2nd Question.
>>
>> I really need to measure 10 MHz to an accuracy of 1 part in 10E-13 / 
>> second (ignoring the accuracy of my workshop standard).  I've tried 
>> multiplying the 10 MHz to the GHz frequencies, but never manage to 
>> improve on the technique I mention in my first question.
>>
>> Any ideas for a relatively cheap way of doing this.  I actually 
>> represent a company that can measure to parts in 10E-15 in one 
>> second, but their boxes costs $50k.
>>
>> Best Regards
>>
>> Martyn
>>
>>   
> Martyn
>
> To achieve this you need:
>
> 1) A frequency standard that has an ADEV < 1E-13 for Tau = 1 sec.
> Such sources tend to be rare and expensive. (eg hydrogen maser, BVA 
> OCXO etc).
> The frequency of the standard also has to be known to better than 1E-13.
>
> 2) Heterodyne techniques can easily achieve the required resolution 
> but achieving the low noise and drift isnt easy:
>      a) The zerocrossing detector shaping the mixer beat frequency 
> output has to be designed to amplify the slope whilst minimising the 
> noise.
>    The naive design approaches often advocated are inadequate.
>    Low frequency ground loops can easily corrupt the measurements.
>      b) The temperature of the mixer has to be held constant to better 
> than 0.005C.
>
>    c) The zero crossing detector temperature has to be held to 
> constant to better than 1C.
>       A naive zero crossing detector design may require temperature 
> control to 0.001C or better.
>
>    d) A low noise stable offset frequency source of accurately known 
> frequency is required
>
> Even when all of the above conditions are met the measurement range 
> may only be around 0.1Hz or so.
>
> Bruce
>