30-10-2012, 04:19 PM
A NEW TYPE OF BALANCED-BRIDGE CONTROLLED OSCILLATOR
A NEW TYPE OF BALANCED.pdf (Size: 189.16 KB / Downloads: 31)
Controlled Oscillator.pdf (Size: 666.81 KB / Downloads: 67)
ABSTRACT
A novel bridge-stabilized crystal oscillator circuit having
exceptional temperature stability is described. The
contribution to the oscillator temperature coefficient
from the circuit components (exclusive of the crystal) is
reduced to about 10-11/ºC, which is several orders of
magnitude better than conventional oscillator circuits.
This avoids a situation where the overall tempco is limited
by circuit component drift rather than crystal stability,
which can easily occur with conventional circuits
when the crystal is ovenized at a turnover point. Previous
attempts to use a bridge in an oscillator were made
by Meacham, who used an imperfectly balanced bridge,
and Sulzer, who used a balanced pseudo-bridge. The
reasons why these are unsatisfactory are discussed. Although
the bridge greatly reduces reactive frequency
pulling, it does not directly address the additional issue
of pulling due to variations in crystal drive current amplitude.
However, it is an enabling technology for a
novel ALC circuit with greatly improved stability. The
new bridge controlled oscillator is also much less sensitive
to other environmental effects such as humidity,
power supply voltage, load impedance, and stray capacitance.
Introduction
This paper begins with a review of conventional
oscillator technology, with particular attention to temperature
sensitivity. Next, the bridge stabilized oscillators
of Meacham and Sulzer are discussed. This will
establish why a new bridge stabilized circuit is needed,
what it needs to do, and how it might work. Then the
new bridge network and ancillary circuits are described.
Within this context, the ALC techniques are developed.
Some manufacturability issues and experimental results
are presented. The oven and other peripheral hardware
used with the oscillator are described in [1] .
Conventional circuits
Virtually all non-bridge-stabilized crystal oscillators
are derived from the free-running Colpitts oscillator.
In the Pierce oscillator [2], a parallel resonant crystal,
replaces the inductor. In the bridged-T1 (also known
as a grounded-base or Butler2) oscillator, a series resonant
crystal is inserted in series between the emitter and
the tank capacitors. (Fig. 1).
The Meacham Bridge Oscillator
The first known attempt to eliminate active device
tempco in crystal oscillators was the bridgestabilized
circuit described by Meacham in 1938 [4, 7] .
In the Meacham bridge oscillator, the crystal replaces
one of the resistors in a Wheatstone bridge (fig 4). The
bridge is connected so as to provide negative feedback to
the active device at all frequencies except a narrow band
around the crystal’s series resonant frequency, where the
feedback becomes positive. Note that the bridge must be
slightly out of balance at the frequency of oscillation; if it
were balanced there would be no feedback to sustain oscillation.
Bridge servo’ed oscillators
In the early 1950’s, there were experiments with
stabilizing a crystal oscillator by using an auxiliary
crystal in a bridge circuit operating in balance. [9] The
idea was that a servo loop would tune the oscillator to
keep the bridge balanced, which would force the frequency
of oscillator to coincide with the resonant frequency
of the crystal at all times (fig. 6). The bridge was
the same modified Wheatstone bridge used in the
Meacham oscillator, and the servo loop was a conventional
lock-in amplifier operating via synchronously detected
audio modulation as commonly used in atomic
standards. The advantage of this scheme is that the
bridge could operate with arbitrarily small deviation
from balance and there was no limit to the servo loop
gain. The obvious disadvantage was the requirement for
two crystals.
A true balanced-bridge controlled oscillator
The approach chosen to meet these requirements
is shown in fig 8. The crystal is removed from a
VCXO and embedded in a two port bridge network.
Meanwhile, port 1 of the network is connected to the
VCXO in place of the crystal. A null detector operates
an AFC (automatic frequency control) loop. The key is
to devise a two-port bridge network that meets two requirements:
(1) The transfer function from one port to
the other has a null output when the crystal is excited at
resonance (as in the modified Wheatstone bridge) and (2)
Port 1, the input, has a driving point impedance such
that it emulates the impedance of a freestanding crystal
(i.e. the bridge is “transparent”). Transparency, as used
here, means that if the oscillator were connected to port 1
of the network, it would operate just as if it were connected
directly to a crystal. In other words, port 1 emulates
a virtual crystal.
Automatic level control (ALC)
In a precision crystal oscillator, the crystal drive
current must be very well stabilized, because the crystal
exhibits a significant amplitude to frequency (AM/FM)
conversion coefficient due to nonlinearities inherent in
the quartz. Fig. 12 shows the AM/FM curve, as measured
in the oscillator described in this paper, by adjusting
the ALC set point. The design goal was to hold the drive
current stable to 0.0002 dB/°C, resulting in a tempco
contribution of about 10-12/°C.
Oscillator gain control
The oscillator must have a gain control input to
connect to the ALC integrator. Typically, this is done by
varying the oscillator collector current. While it is
clearly the simplest method, it, requires that the oscillator
run in a starved bias condition that increases distortion,
and it is difficult to get a lot of dynamic range while
keeping collector current within reasonable limits. Another
problem is that collector current tends to get out of
control at turn on, and can cause oscillator starting
problems. In a bridge controlled oscillator, extra dynamic
range in the ALC is needed to maintain oscillation
during AFC acquisition.
Bridge balance issues
The problem of balancing the bridge involves
two degrees of freedom: The crystal must be at series
resonance to remove any reactive imbalance, and the
image resistor must match the ESR of the crystal to remove
any resistive imbalance. In practice, the AFC loop
has no difficulty servo’ing out the reactive imbalance by
controlling the frequency. The resistive imbalance must
be dealt with on an open loop basis. The ESR of crystals
varies somewhat from unit to unit and is also has a moderate
dependence on temperature and even a slight dependence
on drive level. As a practical matter, then, the
image resistor needs to be adjustable so that the bridges
in oscillators can be resistively balanced on an individual
basis. This adjustment is made by setting the image resistor
adjustment such that the signal at the RF input of
the multiplier reaches a null. This null must at least be
good enough to avoid exceeding the dynamic range of
the multiplier. It becomes more critical as the gain of the
RF amplifier increases, which puts a practical limit on
gain.
Conclusions
A practical technique for controlling the frequency
of a crystal oscillator with a bridge has been described.
The high performance of this technique has
been demonstrated. Critical design issues have been
outlined along with methods of resolving them.