[sdiy] Measuring phase diff between sines

[Ian Fritz]

Fredrik --

Ye Olde LabRat would look at using an accurate DSP lockin amplifier set at 
quadrature.  In this mode the signal out is zero when the waves are in 
phase and increases with the phase difference.  The SR830 claims phase 
resolution to .01deg.  Look here under "digital phase shifting".
http://www.korins.com/m/srs/body_sr830.html

Offhand, I would guess you still need a fairly large time constant, but 
that depends on your noise.

   Ian


At 08:10 AM 5/9/06, Fredrik Carlqvist wrote:

>I have two sine sources and wish to measure the phase difference with at
>least 14 bits accuracy, preferrably 16 bits. The frequency of one of
>these sources is fixed at 10kHz, the other comes from an inductive
>pickup. The maximum phase difference change in one period of the
>reference is 3.6 degrees, that is 1% of a period.
>
>Directly measuring the time between zero-transitions is not good enough
>since there is also noise and DC offset on the picked up signal.
>
>My first thought was to use a PLL-type filter, using two comparators to
>get two square signals, followed by some logic (IN1 & !IN2) to derive a
>phase difference in a capacitor. Then use the reference to determine
>which half-plane we're in (leading or trailing?). This is too slow
>though, as the time constant for the RC filter must be very long to
>allow 14-bit precision (1.5 s, 6 s for 16-bit precision).
>
>        ______        ______
>Sine1 |      |______|
>         ______        _____
>Sine2 _|      |______|
>
>Comb  ||____________||_____ into RC filter
>
>
>It would be ideal to remove all amplitude dependence from the system.
>This would be accomplished by only looking at zero transitions. By using
>integrated and differenciated versions of the picked up signal, we could
>have eight zero-transitions per period.
>
>As the maximum change in phase diff is limited, one could use a DCO to
>compare to the picked up signal, keeping guesses of both the next value
>and the derivative of the phase difference, and using the DCO phase to
>compare to the reference signal.
>
>Does anybody have any better ideas?
>
>
>Fredrik C