[time-nuts] DMTD mixer question

Bruce Griffiths bruce.griffiths at xtra.co.nz
Wed Dec 3 19:54:28 UTC 2008


pablo alvarez wrote:
> Hi,
>
> I have been looking at several Dual Mixer Time Difference designs. As
> far as I know the basic architecture is based on generating a tone
> with an small frequency offset respect to the nominal clocks
> frequency, analogue mixing of the clocks, low pass filtering, then a
> slow zero crossing detector and finally a time interval counter or
> time stamp unit.
>
> I have this naïf question: would not it be better and simpler to use
> directly an ECL D flip-flop as a mixer instead of an analogue one? I
> suppose analogue mixers are preferred because they provide with a
> better "differential tempco", but using ECL logic can provide also
> with a good tempco.
>
> What is your experience?
>
>
> Cheers
>
> Pablo
>
>   
Pablo

In principle one could use the metastability characteristics of a D
flipflop to produce high resolution timestamps of the coincidence
between transitions at the D input and the active clock edge.
For example, assemble a long flipflop based  shift register (only the
first few flipflops of which need be ECL or equivalent fast devices and
clock them at 10MHz whilst driving the input flipflop D input at 10MHz
-10Hz.
The phase of the clock advances by 100fs with respect to the D input
ever 10MHz period.
Ideally the probability the the input flipflop will change state
increases smoothly as the D and clock input transitions approach
coincidence.
Thus a high resolution timestamp of the coincidence between the D and
CLK input signal transitions could, in principle, be derived from the
sequence of flipflop transitions captured by the flipflops in the shift
register.
In practice crosstalk and other non ideal effects will reduce the
achievable resolution.

Bruce



More information about the time-nuts mailing list