Email: GoughW@cardiff.ac.uk
CompanyName: Cardiff University, U.K.
Country: U.K.
Abstract: In MIT, a high-frequency
current flows through a source coil. The phase of the voltage induced in a pickup
coil is measured (i) with air separation and (ii) with the sample under investigation
placed between the coils. The phase difference between (i) and (ii) is the quantity
of interest. Typically this is only ~ 1 degree, and it has been estimated that
a measurement accuracy of a few millidegress is necessary. This is a severe
requirement of the stability of the electronics; at 10 MHz, 10 millideg corresponds
to a time of about 3 picoseconds, and light travels less than 1 mm in that time.
A system is described here in which the short-term noise (over a few seconds)
was ~ 1.5 millideg and the drift over ~ 1 day was ~ 10 millideg.
The source coil is a single turn of diameter 8cm, forming part of a resonant
circuit matched to the 50 ohm output of a signal generator at about 7MHz. A
tapping contact on the coil provides a reference signal.
The receiver coil has 5 turns of diameter 6cm, with a pitch 6mm.The received
signal is fed into a preamplifier and then into a phase detector.
Central to the phase detector is an analogue multiplier which multiplies the
signal(V1)by the reference (V2). The chip AD835, rated at 500 MHz, is very suitable.
V1 and V2 are arranged to be approximately in quadrature. This is achived by
passing the reference signal through a few metres of coaxial cable.
The product signal is {(V1 sin(wt+phi)}V2coswt, where phi is the small phase
to be measured. This equals (V1V2/2){sin(2wt+phi)+sin(phi)}. The first term
represents a signal at high frequency, which is filtered out. This leaves the
second term, which is a constant signal proportional to sin(phi), from which
phi can be measured.
It would be difficult to measure a small d.c. signal reliably in the presence
of multiplier offsets, so phase sensitive detection is used. The reference signal
is chopped at 1 kHz, and the resulting chopped d.c. signal is amplified, and
then multiplied by (essentially) a square wave at the chopping frequency. The
resulting signal passes through a simple R-C low-pass filter, giving a d.c.
output proportional to sin(phi), which is very close to phi.
The pre-amplifier has presented difficulty in its design. Unacceptable drift
has proved a problem, especially if the gain is too high. Best results have
been achieved with a two-stage amplifier using EL2030 op-amps, with a gain of
about 4 per stage.
Overall, the system is of low cost, easy to construct and has a performance
which is promising for used in an MIT system.