|
Once radios in the family car
became a standard accessory, a push ensued to make them cheaper, more reliable,
and service-free. A major Achilles' Heel was the high voltage power supply required
to energize vacuum tubes. Known as vibrator power supplies due to using an oscillator
to convert the battery's 12 DC supply (some autos still used 6 V systems) into
AC that could be transformed up to the 300 volts used by most tubes of the day,
most early failures were attributed to the circuits. They also caused annoying noise
in the audio output if careful filtering and installation was not performed. Having
been invented only a couple years earlier, transistors were being designed into
the power supplies, but low-voltage tubes were still needed for the electronics.
In response to the demand, low-voltage tubes were created to fill the gap until
acceptable transistors became available for a fully solid-state radio. This article
discusses some of the problems with low voltage tube design and methods employed
to overcome them. Low Plate-Potential Tubes
By C. E. Atkins
Tung-Sol Electric Incorporated
With the transistor in the output stage, use of these new tubes
in auto radios makes possible a vibratorless supply.
Re-evaluation of design parameters led to this new family of tubes for direct
operation from a 12-volt source. With a compatible power transistor, they make the
elimination of vibrator power supplies in automobile radios possible.
The development of the transistor by Bardeen, Brattain, and Shockley of Bell
Telephone Lab in 1948 started a new phase in the technology of communication and
electronic instrumentation. The rapid development of transistors provides many challenging
opportunities to achieve results heretofore difficult or impossible with vacuum
tubes.
Of immediate interest are those applications where the special advantages of
the transistor can be most fruitfully employed. A clear-cut example is the hearing-aid
field, where the transistor inroad is complete. Another intriguing opportunity is
in automobile radios, and this field received early attention from application engineers.
Many experimental, and a few commercial, all-transistor automobile sets have been
made. They have many good features but are presently expensive and an adequate,
economically practical, supply of the varied transistor types needed for volume
production of such sets is not yet at hand.
Probably the most attractive feature transistors afford in auto-radio design
is the chance to eliminate the vibrator power supply. Aside from its cost and the
jealously husbanded space it requires, this facility is a cause of considerable
field trouble. The car manufacturers rank vibrators No.1 among the components needing
attention during the early period of service. Furthermore, in doing its job, the
vibrator does not suffer in silence. Vibrator noise which is both mechanical and
electrical (in the form of r.f. hash and low-frequency hum) is an irritation to
the set designer. His mastery of this problem, often incomplete, also results in
minor annoyance to the customer.
A single power transistor providing power output levels that are out of the question
with vacuum tubes operating at low anode potentials allows one to discard the vibrator
power supply; provided the functions of amplification, heterodyning, detection,
and automatic gain control can be fulfilled with vacuum tubes at low potentials.
Whether provision of such functions was feasible has been a hotly debated subject.
Tung-Sol took a lively interest in this matter shortly after March 1952 when
the Lamp Department began work on 12-volt sealed-beam lamps. One of the motor car
companies had advised it was changing from 6 volts to 12 volts in its automobiles,
principally to obtain better ignition, but also to save some copper. Later, the
other car companies made the same switch and furthermore adopted the convention
of a negative ground connection. In the past, many makes of automobiles had the
positive terminal grounded. This lack of standardization would have been awkward.
Incidentally, the writer has been unable to rationalize the choice of polarity
made by different car manufacturers, in the past, although he has inquired extensively
of automotive people. One quaint story was to the effect that a certain ground polarity
was required because its opposite caused ground currents returning from the rear
axle via the drive-shaft to de-plate the crank-shaft bearings. In any case, the
trend just noted opened the way for a realistic consideration of a hybrid receiver
employing one power transistor and 4 or 5 vacuum tubes.
Most tube engineers were loathe to embark on a tube program for low plate-potential
service. There was a history of unhappy consequences whenever set designers tried
using tubes at too low an anode potential. For instance, during the war the 6AJ5
was designed in a hurry because the 6AK5 proved unusable in production communication
equipment running at 28 volts, even though this type had been used successfully
in a pilot run. There was an old rule of thumb to the effect that the mu of a tube
must always be less than the plate potential. Operating difficulties were sometimes
encountered with triodes having a mu of 100 operating in a.c.-d.c. sets with anode
potentials of 80 to 100 volts.
If one attempted to use much bias on the tube, the plate current would be cut
off. This sometimes happens in practice. Because of this, some engineers expressed
the offhand opinion that practical tubes at 12 volts Ep must have an
amplification factor, or mu, of less than 10. They felt that, since the grid bias
had to be approximately -1 volt (due to contact potential), there would be no plate
current if the mu were higher than this value. Although the mu decreases rapidly
as the plate voltage is reduced below 20 volts, it is possible to have an appreciable
plate current and Gm under these conditions.
Fig. 1 - Change in mu of triode section of the 12AJ6 as
plate voltage changes.
Fig. 2 - Control-grid resistance (RC1) affects
change of Ip VB heater voltage.
In fact, one is surprised that the Ip is higher than a prediction
based on the tube's characteristics at 100 volts would lead one to expect. From
inspection of the plate families of the type 12AJ6 (a diode-triode) at high and
low plate voltages respectively, it can be seen that the mu, which was 112 at 250
volts, drops to 50 in the neighborhood of 10 volts. Fig. 1 is a dramatic display
of this effect, where the mu of the 12AJ6 is shown to drop from 77 to 50 volts Ep
to 15 at 1.5 volts Ep. The plate current was held constant at the different
values of Ep in taking this data.
It is believed the reason for this phenomenon is that, as the grid bias approaches
zero, the potential minimum which is normally at or very close to the, cathode moves
over in the direction of the grid plane. This changes the effective geometry of
the tube, thus lowering the mu. Also, the initial velocities of the emitted electrons
are more nearly comparable to the accelerating field due to the plate at these low
anode potentials. This undoubtedly modifies the behavior of the tube.
Another serious pitfall was thought to be contact potential. Since this would
establish the bias at which these low-voltage tubes would operate, many engineers
felt that variations of contact potential between different production lots (or
during life in the same tube) would seriously affect the performance of any equipment
using them. Contact potential had been the villain in many an ill-fated marriage
of tube and circuit. However, a deeper insight into the inter-relation between contact
potential and other tube parameters revealed how it might be possible to exploit
the very feature that generated the apprehension.
One facet of this many-sided situation is depicted in Fig. 2. This is a
plot of plate current vs heater voltage in the 12CR6. By shifting the screen voltage
for each value of control-grid resistor, it is possible to adjust Ip
so that it is always .5 ma. with 12 volts on the heater. Hence, all the curves intersect
at this point. Note the dramatic difference in slope between the condition of zero
grid resistor and that where 10 megohms is employed. Even using 1000 ohms makes
a considerable improvement. These phenomena stem from the fact that with higher
heater voltage (and consequent greater cathode temperature), more current flows
to the No. 1 grid because of the greater velocity possessed by the emitted electrons.
Of course, more current will go to the plate and other electrodes as well.
If the resistance in the grid-to-cathode path is low, there can be no change
in grid potential as the grid current increases. If this grid resistance is high,
the grid develops a negative voltage which will increase with increasing cathode
temperature. This rising voltage at the grid will prevent the more energetic electrons
from reaching the plate. Hence, there is a kind of compensation which, as Fig. 2
shows, tends to hold the plate current more nearly constant. In the same manner
Gm and other tube characteristics are leveled out as well. Also, this
compensation affords a measure of balance when the effective cathode area or the
cathode activity changes. At Tung-Sol, several new tube designs have been evolved
embodying this principle.
In the composition of a hybrid auto set, there appeared to be these problems
the tubes must solve:
1. Driving the power output transistor: This step required a shift from voltage
amplification to current amplification, calling for some power output from a vacuum
tube. It was estimated this might be as much as 50 milliwatts.
2. Getting gain: Voltage amplification could use either low-impedance or high-impedance
circuitry. The earliest endeavors took the form of tubes with larger cathodes giving
higher Gm at the price of lower Rp and greater grid-plate
capacitance. After some preliminary experimentation, this approach was temporarily
side-tracked because the conventional tubes with lower Gm but high Rp
and low Cgp appeared to give a quicker answer. Already designed i.f.
transformers could be used, and some early sets were made using standard tubes selected
for the best low-voltage operation. As a matter of fact, while the design of most
of these tubes has undergone considerable revision, one of them, the 12AJ6, remains
unchanged. It is merely a 12AV6 with special processing, tested under the low-voltage
conditions.
3. The a.g.c. problem: The fact that the input signal at the first r.f. grid
is commensurate with the plate voltage posed some new and nasty problems.
4. Stability of gain: Since 12-volt auto sets are expected to function effectively
at supply voltage from 10 to 16 volts inclusive, this apparently heroic requirement
had to be met. Good life is also required under these conditions.
Returning to problem q, the transfer of power, it was felt 50 milliwatts might
be needed to drive the transistor. Whether this will remain so depends, of course,
upon the power gain of available transistors. In some respects, a suitable driver
tube is the keystone of a successful hybrid radio set.
Fig. 3 - Power output in milliwatts of several tube types.
Only the newly developed 12K5 comes close to providing the drive needed by the power
transistor.
As Fig. 3 shows, conventional tubes, even relatively high-power examples,
are inadequate to do the job. Only the newly developed 12K5 comes anywhere near
the power-output objective. The availability of peak current approaching 40 milliamperes
in the vicinity of zero bias - is achieved in the 12K5 by employing the structure
depicted in cross-section in Fig. 4. The plate family and transfer characteristics
obtained with the special design are shown in Fig. 5.
In the 12K5 the space-charge grid principle is employed. This was first disclosed
by Langmuir in 1913. By providing a No.1 grid adjacent to the cathode with a positive
accelerating potential applied to it and a control grid disposed between this accelerator
and the plate, Langmuir was able make tetrodes with very high transconductance.
When screen-grid became commercially available after 1926, sporadic efforts were
made to this principle in equipment. However, most of the tetrodes and pentodes
available to the equipment designer did not operate well under space-charge conditions.
Furthermore, it was not good engineering economics to provide the power required
by the space-charge grid, as the gain in transconductance did not generally warrant
the cost entailed in exploiting it.
But in a car radio there is a different kind of logic. Since the voltage is low,
it is easily provided by the car-carried battery. The 50 or 100 ma. required for
the space-charge grid is minuscule in the overall current needs of the modern automobile.
Accordingly, this ancient stratagem, long discarded, was revived for the tube that
became the 12K5. The geometry of this type provides a fine-pitch No. 1 grid having
150 turns-per-inch spaced .018" from the cathode, which is fairly large, with an
area of 1.8 square centimeters. The grid-to-cathode spacing is generous and poses
no manufacturing difficulties. The No. 2 or control grid, having 80 turns-per-inch,
is a bit closer to both No. 1 grid and plate. Here the spacing is approximately
.012" each way. Holding this configuration is a bit tricky in production.
This disposition of elements provides the desired features. It is believed that
the fine No.1 grid to which the 10- to 16-volt car-battery potential is applied
accelerates the electrons and groups them into thin sheaths. It is well known that
the factor that limits current in thermionic devices is the repelling effect of
the electrons upon one another when they are crowded together. The provision of
a multitude of layers where the space-charge density is low probably helps to achieve
high space currents with low applied potentials. In addition to its power-output
performance, the 12K5 is an excellent low-voltage relay control tube. It is possible
to achieve 8 or 10 ma. of plate current with only 2 volts of plate potential.
As mentioned earlier, providing adequate automatic gain control was a real problem.
The usual a.g.c. potential has a magnitude comparable to the anode potential of
the i.f. amplifier. This is insufficient to protect the No.·1 grid of the first
stage against positive excursions on very strong signals. Accordingly, there is
bound to be spill-over through the first stage, and this can be sufficient to overload
the converter tube. However, designers hit upon the expedient of applying additional
a.g.c. voltage to the suppressor grid of the r.f. tube in order to limit the amount
of signal getting to the following stage. To do this successfully, certain compromises
have been forced on the tube designer.
Fig. 4 - Location of special space-charge or accelerating
grid in 12K5 tetrode.
Fig 5 - Characteristics of the 12K5 obtained with construction
shown in Fig. 4.
Fig. 6 - Effect of aging on tube characteristics, using
type 12AD6 as an example. Characteristics of partially aged tube is shown in (A),
of fully aged tube in (B)
With high impedance i.f. transformers, it is necessary that the amplifier tube
have a correspondingly high dynamic plate resistance (Rp). This has been
difficult to achieve in tubes operating at low anode potentials. The beam-forming
plates and suppressor grids, which are effective in high potential devices, do not
give the same favorable results with operating potentials in the region of 10 to
15 volts. In fact, the suppressor grids employed in these hybrid r.f. tubes are
probably more useful in reducing the control-grid-to-plate capacity than they are
in providing effective suppression.
The Rp of a tetrode is considerably influenced by secondary emission
from the anode. These secondary electrons leave the plate and are collected by the
screen grid. Since the magnitude of this effect is sensitive to the applied potentials,
a condition exists whereby small increments of anode potential can cause either
a drop in plate current or too rapid a rise in plate current. In the latter case,
Rp will be too low whereas, in the former case, it will be negative.
Either of these effects can be harmful. On many tubes it was found that a nice balance
between negative and positive Rp along the plate volt-ampere curve resulted
in Rp values which were astonishingly high (several megohms where, without
this effect, the Rp would have been a few hundred-thousand ohms).
This dip in the characteristic is shown in the curves for the 12AD6 in Fig. 6A.
Those readers who are old enough to remember the early screen-grid tetrodes (type
24A), will recognize this shape as the dynatron kink. Pentodes were introduced to
cope with this characteristic. A coarse-pitch No. 3 grid placed between the screen
grid and the plate served to thwart the return of electrons from plate to screen.
Because of the geometry involved, this grid did not adversely affect the flow of
primary electrons to the plate to any great extent. As stated earlier, suppressor-grid
techniques do not function as well with tubes at low voltages. While suppressor
grids are still used in the hybrid-set tube designs, it is mainly to help reduce
the grid-plate capacity and because this grid is sometimes needed to secure adequate
a.g.c.
It was found necessary to rely on special materials and processing in order to
iron out the plate characteristic and secure the needed control over the Rp
The efficacy of these measures is greatly facilitated by the important fact that
these tubes are used at very low voltages, so that there is very little plate or
screen dissipation. The heat developed by the electrodes in more conventional operation
has the effect of disturbing or destroying any surface film left on the electrodes
by special processing. In contrast, it is possible to obtain stability and long
life at low voltages with special surfaces too fragile to long endure under the
conditions of high voltage operation. Fig. 6B shows the change in the plate
characteristic of the same tube depicted in Fig. 6A after it received special
aging.
To gain a partial understanding of these phenomena, reference is made to an article
by Matheson and Nergaard in the RCA Review, June 1951, entitled "High-Speed Ten-Volt
Effect." This is a study of the behavior of anode currents in thermionic diodes
with oxide-coated cathodes, operating with plate voltages in the region of 10 volts.
In an ideal space-charge limited diode the current is proportional to the 3/2 power
of the applied voltage. It was noted, however, that in the neighborhood of 10 volts,
there was a small deviation from this 3/2 power. This was found only in diodes with
an oxide-coated cathode. Experiments performed to discover the cause in the coating
itself gave negative results, so a study was made of the anode.
An experiment was performed wherein an anode which had not previously been exposed
to the cathode was rotated in place of one that had been in position when the cathode
was broken down and activated. This new anode did not exhibit the 10-volt effect
until after some time. The effect was then explained by the fact that a layer of
barium from the hot cathode had condensed on the anode, thus giving rise to secondary
emission. This phenomenon is, of course, applicable to triodes and tetrodes - or
pentodes. The effect is manifested by the erratic change of Rp as the
critical plate voltage is approached.
From the foregoing, one might think that this reflection of electrons - if such
it is - might be eliminated by providing the anode with an especially clean naked
surface. This has not proved feasible because it is hard to long maintain such surfaces
in operating vacuum tubes. It has so far not been possible to make good tubes with
bright nickel plates. If, however, carbonized nickel is used and the device is processed
in such a way that there is probably a layer of cathode constituents on top of this
carbon coating, a composite surface is achieved which seems to be quite effective
in curtailing reflected electrons.
This situation is analogous to that obtained when an optical lens or prism is
coated with a thin film in order to increase light transmission by reducing reflection
from the surface. To comprehend this parallel, you must recall that deBroglie in
1924 advanced the theory that electrons had wave properties as well as the particle
characteristics assigned to them. This was verified experimentally by Davisson and
Germer in 1927. According to this theory, electrons have an associated wavelength
which varies inversely as the speed with which they are traveling.
Since 10-volt electrons proceed rather slowly, their wavelength is rather long
as such things go. This wavelength is on the order of 3 or 4 angstrom units. This
is of the same relative magnitude as the film thickness discussed in connection
with the plate. With this state of affairs, it is credible that there is a relation
between the deBroglie wavelength of the slow-moving electrons and the refraction
characteristics of the various media these electrons move through when they impinge
on the plate. The writer believes some such machinery operates to curtail reflection
which, in turn, makes possible better control of Rp at these low operating
voltages.
Thus, by using processing techniques not generally suited to vacuum tubes intended
for conventional higher-voltage service, and by altering the electrode geometry
to take advantage of contact-potential bias, Tung-Sol's engineers have designed
a set of tubes that effectively perform with all electrodes energized by the 12-volt
car battery. With a suitable power transistor, this makes possible an automobile
radio without a vibrator power supply.
These tubes are not final in any sense. The writer believes many more and better
types will be developed by the tube industry for this class of service before the
day of the all-transistor car radio arrives.
In the development of these tubes, many engineers made worthwhile contributions.
The program owes a special debt to Mr. Fred Crawford of the Tung-Sol design department
for many important contributions, especially in the field of tube processing.
Posted December 29, 2021
(updated from original post on 8/20/2013)
|
