amplifiers (aka "mag amps") use a property of saturated core inductors
(saturable reactors) to obtain signal amplification via a transformer-type
plus diode-assisted voltage multiplication. Magnetic amplifiers were
preferred over vacuum tube amplifiers in some circuits because they
do not require a high bias voltage, are generally smaller in size, are
quite robust and are practically immune to microphonics. Their biggest
limitation is bandwidth. The high number of turns in the core provides
a lot of interwinding capacitance so the self-resonant frequency is
in the low megahertz range. Additionally, the need for the magnetic
amplifier reactor to operate in a saturated condition further imposed
a limit on the frequency response. Even today, there are some critical
applications that exploit the fool-proof and ultra high reliability
nature of the magnetic amplifier. This article goes into the details
of operation both as amplifiers and as bistable multivibrators.
Subminiature Magnetic Amplifiers
By A. Hugh Argabrite
Patent Engineer Berkeley Div., Beckman Instruments,
bobbin of wire used in the Ferristor, shown beside vacuum tube for comparison.
New "Ferristor," small, ultra-fast saturable reactor, has high
reliability for industrial automation needs.
The demand for
reliability in electronic equipment is constantly increasing. This demand
has revived interest in one of the earliest practical amplifiers of
electric signals - the magnetic amplifier.
One of the first
models was built by the famous American radio pioneer, E. F. W. Alexanderson,
over forty years ago. By 1945 the magnetic amplifier had proven itself
dependable by years of service in military and industrial control applications,
but its usefulness was limited because of its long response time. This
slowness was due to the necessarily low "carrier" frequencies. The magnetic
amplifier "carrier" is basically the same as the carrier used in radio
broadcasting. A relatively high frequency wave is modulated by or "carries"
the lower frequency "intelligence" or desired signal.
losses had precluded the use of high carrier frequencies. But now, development
of the new low-loss, high-permeability ferromagnetic alloys has made
magnetic amplifiers practical for electronic circuits. With the reduction
of core losses, carrier frequencies up to 10 megacycles are possible.
This gives magnetic amplifiers the fast response required in modern
One of these new units is the "Ferristor"
saturable reactor, made by the Berkeley Division of Beckman Instruments,
Inc. It is a tiny device, not much larger than the eraser on a pencil.
Technically it is a subminiature, ultrafast saturable reactor.
Fig. 1. Basic magnetic amplifier. Permanent magnet used to provide
Fig. 2. Curves of ferroresonant operation.
Fig. 3. Typical ring counter stage. The varistor is used for
Fig. 4. Typical one-shot multivibrator.
Fig. 5. Circuit of current discriminator.
The "Ferristor" reactor is very simple in construction. It has two solenoid-type
windings on a magnetic core. The inner, or carrier-current, winding
has a few hundred turns of very fine copper wire. The outer, or control,
winding has from one thousand to four thousand turns, depending on its
intended use. The core is a few turns of sheet magnetic material, 1/8000
of an inch thick, rolled into a scroll. The finished assembly is completely
sealed by "potting" it in epoxy resin.
These units can be used
in at least two fundamentally different ways, either as magnetic amplifiers
or as bi-stable ferroresonant elements. The latter use requires only
one winding however.
When this reactor is used as a magnetic
amplifier, Fig. 1, a small load capacitor is placed in series with the
carrier winding. Changes in control current vary the core permeability.
This changes the effective inductance of the carrier winding. The carrier
current is thus linearly modulated by the waveform of the control current.
The magnitude of the carrier current and the amplitude of the r.f. output
(across the load capacitor) vary in step. The amplified input waveform
is "recovered" by rectifying the output and filtering out the r.f. The
orientation of the rectifying diode determines whether the output is
in- or out-of-phase with the input. The unit amplifies because the carrier-current
variations are much larger than the control current changes. The extreme
sensitivity of magnetic amplifiers enables them to amplify signals weaker
than vacuum-tube shot noise.
The parts of a magnetic amplifier
circuit may be compared to the parts of a vacuum-tube circuit. The control
current corresponds to the control grid voltage; the carrier supply
compares to the "B+" supply; the carrier current is analogous to the
plate current; and the load impedance corresponds to the plate load.
The other basic application of "Ferristor" reactors employs
the phenomenon of ferroresonance to produce a bi-stable characteristic.
With this characteristic they can maintain either of two possible stable
In ferroresonant operation a capacitor
is selected to resonate with the partially saturated carrier winding
inductance near the applied frequency. The resulting circuit has an
S-shaped curve of voltage versus current. See Fig. 2. The capacitive
reactance is constant with increasing current. But when the current
through the iron-cored carrier winding is increased it becomes saturated.
This decreases its effective inductance and the inductive reactance
of the circuit.
Over a limited range, either of two different
stable values of current can exist, for one given voltage. In circuit
operation this voltage is actually the r.f. supply voltage. The intersections
of the curves of power supply output and positive-slope circuit impedance
are the two stable states. In the low current, or non-saturated state,
the circuit is inductively reactive. In the high current, or saturated,
state it is capacitively reactive.
In ferroresonant operation
the control current does not flow constantly. Instead trigger pulses
of control current are used to induce partial saturation of the core.
It can be saturated by an external magnetic field, by d.c. in the control
winding, or by r.f. in the carrier winding (which makes it capable of
A d.c. trigger pulse is applied through the
control winding to raise the carrier current to the high state. The
trigger pulse can be of either polarity, since both will temporarily
reduce inductance. To make it change states, the core must be partially
saturated to the point where any current increase decreases the inductance
enough to drop the output voltage. Thus a pulse causes the current to
re-generatively jump to the stable high current point. The interdependence
of low inductance and high current keeps the circuit latched in this
state of self-saturation. The unit cannot return to the low-current
state until its supply voltage is reduced below the level where self-saturating
current flow starts.
A ferroresonant flip-flop can be constructed
by running two reactors in parallel. They must have a common coupling
element-capacitor or triggering reactor - to the r.f. supply voltage.
If memory and gating circuitry are added, the flip-flop can be operated
with a single input.
An extended version of the flip-flop is
the ring counter, Fig. 3, composed of several reactors operated in a
closed loop. The ring counters are connected to the supply voltage through
a common coupling element. Its value is chosen so only one unit at a
time can draw high current. If two or more try to jump high, the extra
current through the common coupling tends to drop the voltage so neither
can be high.
The action in a ring counter is not random. In
one common circuit, each unit has a directive circuit to transfer the
high current or "active" state from one stage to the next stage in turn
with each successive input pulse. The varistor associated with the "active"
unit has the largest current through it. Thus it will conduct the most
when the ring is pulsed. This increased current also flows in the control
winding of the unit following. The new unit then goes into the high
current state and the previous unit is forced out because of the action
of the common coupling element.
The action of the stages in
a ring counter may be compared with several glow regulator tubes in
parallel across some voltage through a common resistor. As the voltage
is raised, one tube will strike first, lowering the voltage to its regulated
value. This prevents the other tubes from firing since the voltage has
been reduced below their firing potential.
These units can be
used in a one-shot multivibrator circuit. See Fig. 4. This circuit generates
a constant-width output pulse and steepens the leading and trailing
edges of a slow pulse input. The one-shot circuit is made from the magnetic
amplifier circuit by adding a capacitor and resistor in series from
the output to the input. The diode is oriented for in-phase output.
The output pulse width is determined by the discharge time of the resistor
The circuit is triggered by an input pulse which
induces partial saturation. Regeneration causes the carrier current
to jump to the fully saturated state. A change at the output is reflected
back to the input where it adds to the original input signal. This additive
action takes the output practically instantaneously to its saturated
value, where no further change is possible. While current from the capacitor
keeps the core saturated the output stays in the high state. But as
soon as the charge leaks off, the output drops back to the low state.
This circuit can put out 25-volt, 4-microsecond pulses into a 1000-ohm
load. This can be altered with bias changes. Bias can also be used to
cause the circuit to trigger at a particular point on the input wave.
The circuit shown requires a 15-volt positive pulse drive and puts out
a pulse 40-microseconds wide.
A current discriminator, Fig.
5, which generates rapidly changing waveforms from slowly varying input
waveforms, can also be made from these reactors. It is the analogue
of the familiar vacuum-tube voltage discriminator, or the Schmitt trigger
Ferristorized decimal counting unit (right) compared with the vacuum-tube
To form this circuit positive feedback is added to a magnetic
amplifier through a series resistor from the output to the input. The
diode rectifier is oriented for in-phase output. The discriminator has
two possible stable states, either high- or low-current output. When
the input current is above a certain value the output will be saturated.
When the input current is below another certain value the output current
is low and non-saturated. The circuit "hysteresis" is the difference
between these two input current values. The magnitude of the hysteresis
depends on the size of the feedback resistance. When the value of input
current is within the range of these values the output state depends
on whether the input was previously above or below the "hysteresis"
range. The change of state is caused by regeneration. In the circuit
shown the hysteresis is 130 microamperes.
The current discriminator
may also be used as a pulse gate. A d.c. bias fixes the static level
of the control current without input pulses. If the d.c. level is high
the gate is "open," because input pulses can reach the region of partial
saturation to produce output pulses. If the d.c. level is low the gate
"Ferristor" reactors are very versatile. They can
be used in many circuits including oscillators, multivibrators, discriminators,
balanced or differential amplifiers, coincidence gates, ring counters,
binary counters, stepping registers, and many others.
reactors are almost completely unaffected by shock, vibration, temperature,
humidity, altitude, time, and the other things affecting vacuum tubes.
They are reactive rather than dissipative devices so they generate little
heat and there is practically no power waste. For these reasons they
should last almost indefinitely.
The invention of the audion
displaced the original magnetic amplifier from favor. But now its modern
descendant - the "Ferristor" reactor - with its greater inherent ruggedness
may replace the relatively short-lived vacuum tube for certain applications
requiring good reliability.
Posted January 5, 2014