[time-nuts] Question about precise frequency / phase measurement

Magnus Danielson magnus at rubidium.dyndns.org
Sat Apr 21 04:22:49 EDT 2012

On 04/20/2012 03:49 PM, EWKehren at aol.com wrote:
> It is a circuit that they for instance use in the 2110 where they take the
> reference input in case of 10 MHz divide by 2 and also divide the 5 MHz
> down to  500 Hz use an exor and out comes 5.000500 MHz filtered and divided by
> 5.

A similar approach is being used in many rubidiums to generate the 
5,3125 MHz signal, taking the 5 MHz, divide it by 16 and then XOR them 
together and then toss it through a LC tank to pick out the right 
frequency component. The rubidium is then tweaked using the C-field such 
that the locked 5 MHz lines up with SI second. That's just one aspect of 
why rubidiums have been relatively cheap from the start. The isotopic 
match of D lines allowing fairly easy filtering and selective pumping is 

> The  result is 1.000100 MHz which is mixed with the unknown divided to 1
> MHz. The  result is 100 Hz counted with a 100 MHz period counter and you
> have 1 E 12 in a  second. My counter which is part of the system and thanks to
> Richard MCC is a  PIC, has 0.1, 1, 10 and 100 seconds gate time. The 100 or
> 200 MHz are generated  from the reference channel. All logic is in a
> MAX3000A G/A. The output is RS 232  and can also be stored on a USB memory stick,
> no PC needed. Austron uses a Xtal  as a filter and I use 2 because I do not
> have access to their Xtal but it works.  A nicer solution would be to use an
> AD 9833 DDS but it would require an   additional PIC to do the math since
> the DDS can not produce an exact 1.000100  MHz. If some one is willing to do
> that chip please contact me off list.

If you do quadrature signal multiplication, you can avoid the mirror 
frequency without high-Q filters. Look up the Tayloe detector for some 

In this case you can generate an I and Q signal by adding a DFF. By 
producing a 2 kHz and 500 Hz, you let the 500 Hz be the I signal and 
then let the additional DFF have that as D input and clocked by the 2 
KHz it will produce the 90 degree shifted Q signal on the Q output.

By quadrature separation of the 5 MHz you can then use the 5 MHz I and 
Q, mix and then analogue sum prior some mild filtering such as a LC-tank.


More information about the time-nuts mailing list