June 1969 Electronics World
Table of Contents
Wax nostalgic about and learn from the history of early electronics. See articles
Electronics World, published May 1959
- December 1971. All copyrights hereby acknowledged.
This edition of Electronics World ran a series of diode articles:
Hot Carrier Diodes,
Microwave Power Diodes,
A Survey of
Silicon Junction Diodes, and
Microwave Power Diodes
Dr. Niblack received his B.A. from the University of Buffalo, his A.M. in
Physics from Harvard, and his Ph.D. from State University of New York. He joined
Sylvania in 1961 and for three years carried out analysis of antennas and receiver
systems for high frequency, microwave, and optical use and worked on the effects
of radiation in semiconductors. From 1965 to 1967 he was with Motorola where
he developed designs and processing for advanced MOS devices in power transistors.
He rejoined Sylvania in the Semiconductor Division in 1967. He taught physics
at two universities.
Mr. Levi holds a B.A. in chemistry from University of Southern California.
After two years of graduate work in chemistry, he joined Pacific Semiconductor.
He has also been associated with Clevite Transistor Co. and Raytheon. He joined
the staff of the Microwave Division of Sylvania in 1967. He holds several patents
in the field.
By Walter K. Niblack and Clifford A. Levi
Microwave Dept., Semiconductor Div., Sylvania Electric Products
New avalanche and transferred-electron (bulk) devices, such as Read-effect
or IMPATT, TRAPATT, Gunn-effect, and LSA diodes, are making possible the production.
of microwave power output at frequencies above 50,000 MHz.
Ever since the invention of the transistor, people in the microwave-systems
business have dreamed of being able to produce and modulate microwave power
without the use of vacuum tubes such as klystrons, magnetrons, and traveling-wave
amplifiers. A solid-state transmitter, which shared the advantages of solid-state
circuitry at lower frequencies - chiefly small size, high reliability, high
efficiency, and the elimination of multiple power supplies - was the goal.
The first practical solid-state microwave sources appeared about ten years
ago and, typically, consisted of a v.h.f. transistor oscillator whose output
was fed into several stages of frequency doublers using the nonlinear capacitance
of varactor diodes. At first, these sources were quite inefficient and very
complicated, but had good reliability. Improvements in transistor and varactor
design allowed the use of one v.h.f. power transistor, together with a high-order
varactor multiplier (e.g., one whose output frequency is 5 to 15 times the
input frequency). This simplified the circuit considerably, but usually reduced
It was recognized that what was really needed was a device which, like a
klystron or magnetron, would produce output at some microwave frequency with
only d.c. input power, just as a transistor oscillator does at lower frequencies.
Transistors have improved considerably in the past ten years and now there
are some capable of producing 5 watts at 2 GHz; but, at present, these are
hard to make and thus expensive. It appears to be nearly impossible in the
foreseeable future to make transistors that oscillate well above 8 GHz because
of basic limitations.
In the past two years, however, two new classes of diode devices have emerged
which make it unnecessary to use transistors at all in many oscillator and
amplifier applications above 1 GHz. These are the avalanche-type devices which
may be used either as oscillators or as negative-resistance amplifiers; and
the transferred-electron devices which so far have found practical application
only as oscillators. Within each class there are a number of different "modes"
or types of operation possible. Although it is usually necessary to optimize
a given device design according to the intended mode and frequency desired,
a particular device can often be operated in several different modes as circuit
conditions and d.c. input are varied.
Devices in both these classes have produced power at frequencies above 50
GHz, far above the probable upper limit for transistors. Being two-terminal
devices, they require only very simple d.c. power supplies and no bias on a
third terminal. Also, both types will oscillate in the transit-time mode in
almost any high-"Q" cavity if efficiency is not important.
There is some confusion about the modes since various manufacturers have
given different names to the same mode. For the avalanche devices, there are
three recognized modes: (1) the Read-effect, or transit time, or IMPATT (IMPact
ionization And Transit Time) mode; (2) the "anomolous," or subharmonic, or
TRAPATT (TRApped Plasma And Transit Time) mode; and (3) the self-pumped parametric
mode. There are four clearly recognized modes for transferred-electron (bulk)
devices: (1) the Gunn-effect, or transit-time mode, (2) the delayed-domain
mode, (3) the LSA (Limited Space-charge Accumulation) mode, and (4) the quenched
multiple-domain mode. Since both types of device exhibit complicated behavior,
it is possible that other modes will be identified in the future.
With solid-state sources available to convert d.c. directly to microwave
r.f., some means is required to modulate the output. Simply varying the d.c.
input will do the trick, but in optimized circuits this results in simultaneous
changes in both frequency and amplitude, so other means are desirable. If frequency
modulation is desired, the best solution is to apply the modulating voltage
to a tuning varactor coupled to the oscillator circuit. (These diodes are covered
in another article in this Special Section.)
If amplitude modulation or pulse modulation is desired, it has been the
practice for a number of years to use p-i-n diodes. These diodes can be used
as current-variable attenuators in forward bias, but their primary use is as
switches for pulse modulation and phase shifting. As high-speed switches, they
can also handle very large amounts of incident power (up to 10 kW pulsed power
and hundreds of watts c.w.) because of their low loss in both the "on" and
"off" states, and high breakdown voltage.
Fig. 1. (A) Cross-section schematic diagram and (B) typical impurity profile
of an avalanche-type diode.
Fig. 2. By using diamond as part of the heat sink, the thermal performance
of avalanche diode is improved.
Table 1. Partial listing of some available IMPATT diodes.
The avalanche IMPATT diode is a two-terminal negative-resistance device
which can be fabricated of silicon, germanium, or gallium arsenide and is shown
in Fig. 1A. The indicated p+ region may be formed by diffusion, epitaxial growth,
or ion implantation; the first being most common.
When the diode is reverse-biased into breakdown, the depleted n zone is
functionally divided into two regions; an avalanche region and a drift region.
In the avalanche region, the field is high enough to cause impact ionization,
which causes continual production of free electrons and holes. This is strongly
dependent on the field within the region. An increase in field causes an increase
in the rate of change in the number of electron-hole pairs, and thus the current.
There is a phase shift between the voltage and current so the avalanche current
acts inductively. A further phase shift occurs across the drift region. If
at some frequency the total phase shift for the two regions is greater than
90°, the diode exhibits a negative resistance at that frequency. This picture
is complicated by the fact that capacitive (displacement) current also flows
in the avalanche and drift regions causing a resonance.
The n+ region serves as a contact for the active region of the diode and
as a support during fabrication of the slice.
Typically, these diodes are mesa structures. An X-band diode might have
a mesa diameter of 3-5 mils and an impurity distribution as shown in Fig. 1B.
Since typical efficiencies in the IMPATT mode are on the order of 10% or
less, an input power of over 10 watts is required for 1-watt output, at say
10 GHz. At least 9 watts must therefore be dissipated. Hence, the thermal design
is extremely important. For this reason, avalanche diodes are mounted with
the anode bonded to a diamond heat sink to reduce the distance from the junction
to the heat sink. Fig. 2 is a drawing of a diode bonded to a diamond heat sink.
Diamond is a better heat conductor than copper by a factor of four and advantage
is taken of this in the critical spreading region of the heat path. Proper
thermal design can permit thermal resistances of less than 6°C/W for a
junction diameter of 5 mils, as opposed to typically 15°C/W for the chip
mounted directly on copper.
In operation, as current is increased, a threshold current is reached, the
value of which depends on diode area, frequency, and doping of the avalanche
region. For a diode which is not thermally limited, the output power then increases
with input power to a saturation value. The efficiency also increases with
input power to near the saturation point. Hence, for maximum efficiency, a
diode must be designed to operate optimally at its design power level.
The performance of an IMPATT diode depends critically upon the design of
the circuit in which it is operated. Efficient operation has been achieved
in waveguide cavities, coaxial cavities, stripline, and microstrip circuits.
In the design of a circuit, attention must not only be given to the impedance
presented to the diode at the fundamental frequency, but, in oscillator applications,
to the proper terminations of the harmonics.
In the laboratory, power output at X-band of approximately 2 watts c.w.
has been obtained in a single-chip device operating as an oscillator in the
IMPATT mode. Similarly, multiple-chip devices have been built which generate
as much as 5 watts c.w. Frequencies covered have ranged from 500 MHz to 70
GHz with 100-mW output being achieved in the laboratory at 50 GHz.
Fig. 3. Three-stage avalanche-diode amplifier package.
In the TRAPATT mode, much higher powers and efficiencies have been achieved.
Five watts c.w , at 40% efficiency has been reported for a single chip. In
pulsed operation, much higher powers, on the order of a hundred watts at 40%
or higher efficiency, have been reported. These results are generally at frequencies
below X-band but, in principle, similar efficiencies could be achieved at X-band.
There is little doubt that performance such as that just cited will move out
of the lab and into the market place in the near future.
For the present, both IMPATT diodes and complete oscillator circuits are
commercially available from a number of sources and at power levels up to 1
watt at X-band. Table 1 is a partial listing of available diodes and oscillators.
Since the field is moving so rapidly, it is difficult to prepare an exhaustive
list. In addition to oscillator circuits, a 3-stage amplifier is available
from Sylvania. This device, shown in the photograph, Fig. 3, provides 40-dB
gain with up to 200-mW output.
Single diode circulator-coupled avalanche amplifiers have been built with
outputs over one watt in the X-band, and bandwidths over 2.5 GHz reported,
using commercial diodes.
Three circuit configurations that have been used for avalanche and transferred-electron
device oscillators, The d.c. power supply should be of the regulated-current
type with avalanche diodes. Voltage, current, and polarity depend on device.
Transferred-Electron (Bulk) Devices
The transferred-electron effect is an effect which occurs because of peculiarities
of the conduction band in certain semiconductors such as gallium arsenide.
It does not occur in semiconductors such as silicon and germanium. The effect
occurs when the field in an n-type sample exceeds a certain threshold, giving
the electrons enough energy to transfer to another conduction band with higher
energy and lower mobility. This results in a decrease of average mobility,
and thus a decrease in conductivity, as the field increases. In simple terms,
bulk gallium arsenide exhibits a negative small-signal resistance above a certain
voltage. A typical I-V curve is shown in Fig 4.
At first it might seem that a slab of gallium arsenide could be made to
amplify, just as a tunnel diode does, simply by connecting it to a positive-resistance
load nearly equal in magnitude to the negative resistance and applying voltage,
and that the frequencv of operation would he determined entirely by a cavity
or a tuned circuit. However, if an excess of electrons is created somewhere,
instead of diffusing away from each other as usual, they diffuse toward each
other leaving behind positively charged donor atoms. This segregation of positive
and negative charges is a "dipole domain."
As soon as ohmic contacts arc made to a piece of gallium arsenide, and sufficient
field is applied, electrons are injected at the negative end and a domain immediately
Fig. 4. Typical I-V curve for bulk gallium arsenide showing the negative
resistance above certain voltage.
Fig. 5. Gunn diode showing the traveling domain produced.
Fig. 6. A practical Gunn diode structure is shown here.
Fig. 7. Operation of "p-i-n" diode with forward-bias.
The electrons in the domain are free to move toward the anode so the whole
domain drifts toward the anode at a velocity of about 107 cm/second.
This is shown in Fig. 5. Since most of the voltage drop is taken up by the
domain, no new domain can form until the existing one is destroyed at the anode,
allowing the bulk field to rise into the negative-resistance region again.
For this reason, the sample oscillates at a natural frequency (f0)
close to 107/L (where L is the cathode-anode spacing in centimeters),
regardless of circuit conditions. In practice, it is possible to "pull" the
oscillation frequency from its natural value, f0 by ±10%
by coupling to an external high-"Q" cavity.
Fabrication of practical Gunn diodes is complicated by the fact that the
value of L must be small (about 10 microns for f0 = 10 GHz) so that
the device cannot be made by the obvious procedure of putting ohmic contacts
on opposite sides of a single crystal slab of thickness L. Epitaxial layers
of lightly doped gallium arsenide on heavily doped substrates must be used.
In addition, the doping must be very light and very uniform if high conversion
efficiency is desired. It is very difficult to meet these conditions, hence
Gunn diodes are expensive to make. The structure of a practical Gunn device
is shown in Fig. 6.
In modes other than the Gunn mode, the creation and annihilation of domains
is controlled by material properties or circuit waveforms. In the LSA mode,
domains are virtually completely suppressed by increasing crystal thickness
and reducing doping level, while improving uniformity (very difficult). Because
gallium arsenide is a poor heat conductor, and high input powers must be used,
LSA devices must be pulsed.
Operating a Gunn diode is extremely simple. The diode is placed in a circuit
that resonates near f0 and d.c. voltage from a low-impedance source
is applied until the field exceeds the critical field, causing domain formation.
As the voltage is increased, the power output increases rapidly until the efficiency
reaches a maximum of 2 to 5 percent. Power output usually continues to increase
with increasing voltage until the device burns out.
A typical X-band Gunn diode might produce 20 to 50 mW output at 10 volts
and 100 mA. The threshold would be about 5 volts, and burnout might occur due
to overheating at as low as 12 volts.
The p-i-n diode, which can be made only in silicon, is basically an electronically
controllable variable resistance. In some applications (switching, phase shifting)
only two of its states are used, that is, "on" and "off." In others, such as
attenuator or modulator, the continuous variation of resistance with forward
current is utilized.
The operation of this diode depends upon the modulation of the conductivity
of an intrinsic (very high resistivity) region by injection of holes and electrons
under forward bias. Fig. 7 is a schematic of a p-i-n diode under forward bias.
At low current levels or at zero or negative bias very few carriers exist in
the i region. Since the conductivity is directly proportional to the number
of carriers, the resistance of this region is high. As the current is increased,
however, the conductivity increases as injected carriers from the p+ and n+
regions dope the i region. This effect is called conductivity modulation and
is exhibited to some extent in all forward-biased p-n junctions when the injected
carrier density becomes comparable to or exceeds the background doping. Fig.
8 is a plot of resistance vs forward current for a typical p-i-n diode. Operating
bias levels for switches or phase shifters might be in the 10-150 mA range.
In the reverse direction, mobile charge is swept out of the i region and
it becomes an insulating dielectric. The diode behaves as a typical reverse-biased
junction with two exceptions. Since the i region has few carriers to be removed,
it is swept out at low voltage and, in a good diode, punches through at a few
volts or occasionally at zero bias. Since the i region is typically fairly
thick, the capacitance per unit area at lower reverse voltages is less than
for a comparable p-n junction.
Fig. 8. Forward-bias characteristics of the "p-i-n" diode.
In summary, then, a p-i-n diode can be represented as a current-variable
resistance in the forward direction and as a constant capacitance in the reverse
In microwave applications, the time during which the applied signal exceeds
the breakdown voltage of the diode is too short for buildup of avalanche current
to occur. A p-i-n diode, may, therefore, control peak r.f. voltage greater
than the breakdown voltage of the diode.
In the forward-biased state, the r.f. current swing can exceed the d.c.
bias without causing much rectification. Since the diode has a finite series
resistance, power is dissipated and, therefore, the thermal design of the diode
is important when appreciable powers are to be handled. This is particularly
important in applications when the diode may not be fully conductivity modulated.
Diodes should be selected with low thermal resistance as one of the criteria.
Table 2. Partial list of suppliers of microwave "p-i-n" diodes along with
typical range of operating parameters.