Application Note – 33
Precision Phase Difference Measurement
Overview
There are many instances in Frequency and Time (and many other) applications where it is convenient to be able to measure the phase difference between two signals.
This paper describes a method for making precision measurements (down to the pico second level) without requiring exotic (or expensive) components. This technique is used within the latest ptf Time and Frequency reference instrument, the ptf 3207A, for determining the phase difference between the internally generated 1PPS (one pulse per second) signal and an externally generated PPS reference, be it from a satellite system, time code, RF or external PPS input.
Principle of Operation
The basic principle of operation uses a combination of a “coarse” phase measurement and a “fine” phase measurement.
The coarse phase measurement consists of using a measurement clock, gated by a start and a stop signal representing the leading edges of the compared signals. These signals are designated as a reference(REF) signal and a unit under test(UUT) signal. Note that for this method to work correctly, the reference signal leading edge must be coincident with the measurement clock leading edge.
Measurement clock pulses are counted between the Start pulse and the Stop pulse. Either the REF or the UUT may start the count, however the calculation is different dependent upon which comes first. For the purposes of this paper we will consider the REF to come first. The coarse phase measurement is simply the number of clock pulses between the Start and Stop signals, however due to the finite resolution of the measurement clock however, there is a residual error remaining which is the result of the difference between when the actual Stop occurred and the Stop was measured by the next edge of the clock. See figure 1 below.
Figure 1.
Expanding the area of interest from figure 1, we can see that for the coarse measurement we would count 7 pulses, see figure 2 below.
The whole objective of the fine phase measurement is to measure the error and calculate the actual position of the Stop pulse. To do this we first generate an error pulse from the actual Stop pulse and the next trailing edge of the measurement clock (the reason for using the trailing edge is to insure we never have a “zero” pulse width – a correction is made to subsequently subtract one clock pulse width). We then use this pulse to drive a precision interpolator, which is analog in nature, and is similar to the diagram shown below, figure 3.
Figure 3.
The error pulse is used to charge a capacitor at a fixed rate (nominally 1V/100ns). Once the error pulse goes low, the count is enabled and the capacitor starts to discharge at a rate 1000 times slower than the charging rate.
The number of clock pulses measured is therefore 1000 times bigger than the coarse measure mode, resulting in an effective error amplification of 1000. With a 10MHz clock, this gives an attainable resolution of 1000 times a 100ns pulse, i.e. 100 pico seconds.
In practice, it is necessary to calibrate the precision interpolator at regular intervals, to compensate for any variance in the analog components with time/temperature etc.
The calibration method shown below is for a system using a 10MHz measurement clock.
Measure the count for a 150nsec cal pulse width
Measure the count for a 50nsec cal pulse width
For a negative sign (as per our description, REF first) calculate the slope:
m = (c1 – c2) /(100)
solve for the y intercept:
b = c1 – (100)*m
b = c2
y = m*x +b where y is in counts and x is in nsec
To calculate the time interval from counts, solve for x in terms of y:
x = (y – b)/m ideally: x = (1500 – y)/10
terror(nsec) = 100(c2 – counts)/(c2 – c1)
Summary
This technique provides a highly effective way of making precision phase measurements without the need for exotic (and expensive) components. The method is used with great success within the ptf 3207A range of GNS receivers and allows phase measurements to be made with a resolution of 100 pico seconds, resulting in a much smother control loop and subsequent improved frequency and timing accuracy and precision.