Understanding Diodes & RF/Microwave Operation

Brooke Clarke 2000 - 2015



Two Types of Diodes
    Charge Storage
    No Charge Storage
Microwave Mixers
Links

Two Types of Diodes

It is very important to understand the difference between charge storage and non charge storage diodes.  This is a fundamental distinction.

Diodes That Store Charge

These include but  are not limited to: PIN (Wiki), Varactor (Wiki), Step Recovery Diodes (Wiki SRD) and Noise diodes (Wiki).  In some cases a guard ring Schottky (Wiki) that is improperly constructed can exhibit charge storage.

If there is charge storage then there is a corner frequency [Fcs = 1/(charge storage time)] and if the signal of interest is well above Fcs then the operating point of the diode does not follow the I-V curve.  The RF signal sees an almost constant diode impedance.  A PIN diode can be used to switch high powers for frequencies that are well above Fcs.  The diode capacitance at low frequencies is different from the capacitance well above Fcs where it is the same as the depleted (large reverse bias) low frequency capacitance (Wiki) (even with no bias).

It is important to characterize a charge storage diode both above and below Fcs as well as knowing the storage time (or lifetime).  If this is not done, then diodes that have some (but not all) parameters the same will NOT work the same in the final application.

Also see my Radar Warning Receivers web page for detectors, limiters and limiter-detectors and the AM-6536 / ALR-54 Radar Warning Receiver Front-end for another example.

SRD

aka Snap Diode.  When driven in the forward direction the diode accumulates charge in the Intrinsic region.  When the polarity is reversed a diode normally acts like an open circuit, but not in this case where the diode conducts in the reverse direction until the stored charged has been removed.  At that point the diode quickly (transition time) turns off.  So to maximize the voltage when the diode turns off the life time should be chosen to be 1/4 of the period of an input sine wave signal.  Or if a pulse type signal is being used the forward pulse should be long enough to charge the diode and the reverse portion should be longer than the lifetime.

The lifetime increases with temperature.  That's why HP made a component combining a SRD, heater and temperature sensor for the 5100 Frequency Synthesizer.  The input frequency is 1 MHz i.e. period is 1 micro second so the diode should have a lifetime of about 250 ns.  Microwave type SRDs typically have much shorter lifetimes.

HP Journal Dec 1964 -
3374416 Impulse circuit including a step-recovery diode, Hall Robert D, Krakauer Stewart M, (Hewlett Packard Co) Mar 19, 1968, 363/158, 327/303

PIN

These look like an open or short.  A simple switch could be made by connecting the PIN diode between a center conductor and ground.  When off (no or back bias) a signal would just pass the diode, but when forward biased the diode would look like a short stopping the signal but also reflecting all of it.  It would be better if the switch was matched to the transmission line in both states. The following patent shows some ways to do that:

3245014 Microwave switch, Hall Robert D, Hyman Plutchok, Apr 5, 1966, 333/262, 327/493 -
3503014 Multiple throw microwave switch, Hall Robert D, Nelson William W, (Hewlett Packard Co, Mar 24, 1970, 333/104, 327/504, 333/207

Diodes That do Not Store Charge

These include Schottky (Wiki)  (both simple and guard ring, Tunnel (aka Back) diodes(Wiki).  In this type of diode the RF signal follows the I-V curve.  For example a Schottky diode can be used as a microwave detector but can not handle high power.

Schottky

This type of diode also called a "hot carrier" diode is formed with a metal-semiconductor junction.  They can be made using either P or N type semiconductors and with different metals resulting in different barrier heights.  For diodes with about the same area the barrier height shows up as a difference in the forward voltage at 1 ma.  All of these simple Schottky diodes have a reverse break down voltage in the 3 to 9 volt range.

Massive Schottky diodes can have very low forward voltage drops.

Space Charge Limited Current

There is a limit to the current that can flow through a Schottky diode.  It is defined in Sze's book "Physics of Semiconductor Devices".  When Schottky diodes are used in mixers the mixer performance has limitations because of space charge current limiting.  A measure of this effect can be had by looking at the diode resistance as measured by a 100 ma current pulse that is very narrow.  I remember heavily funded government programs trying to get a handle on this effect but not knowing what the effect was.  It may still be a mystery to many people.
WJ App Note: Predicting intermodulation (IM) suppressionin double-balanced (DB) mixers - did not understand this.
WJ App Note: Mixers in Microwave Sysems Part 1 - did not understand this.
WJ App Note: Mixers in Microwave Sysems Part 2 - did not understand this.

Guard Ring Schottky

This is a Schottky diode with two additional diode junctions added.  There is a "guard ring" junction surrounding the normal Schottky junction that also goes under the normal junction.  If it is done correctly the guard ring increases the reverse breakdown voltage to the 30 - 45 volt range and the third junction never turns on in the forward direction.  If the third junction does turn on then it can store charge and degrade the performance of the diode.  To check for this the forward I-V curve is inspected to see if there is a "double break".  The forward curve should have a single smooth exponential rise.

3463971 (Google) Hybrid Semiconductor Device Including Diffused-junction and Schottky-barrier Diodes, R.W. Soshea, R.A. Xettler (HP), Aug 26 1969
3463971 Hybrid
          Semiconductor Device Including Diffused-junction and
          schottky-barrier Diodes
Examples of guard ring diodes are the 1N5711, 1N5712, 5082-2800, 5082-2810, 5082-2811, 5082-2835 currently offered by Avago.

Automated Testing

The DC behavior can be well modeled by assuming the diode follows the classical exponential I-V curve and in series has a fixed series resistance (Rs).  By measuring a few points on the I-V curve and doing some curve fitting the Rs can be separated from the diode junction.

When measuring these low diode voltages a Kelvin connection is needed for good accuracy.

Tunnel & Back Diodes

Aertech, which started in Mtn. View, CA then moved to Sunnyvale, CA.  Along the way they were bought by TRW and later by FEI, and finally went out of business when the cold war ended.  Metelics was founded by someone who worked at Aertech.

Aertech had two key product lines:

Back Diode detectors and Tunnel diode amplifiers.  I worked mainly on the TDAs, but knew most of what went on with the detectors.  A Back Diode is a tunnel that has a peak current less than 1 mA and is designed for use in detectors.

I built test equipment for characterizing Tunnel (and back) diodes.

The impedance that the diode "sees" looking back into the test system MUST be less than the -R value of the diode AT ALL FREQUENCIES less than the self resonate frequency of the diode.  For tunnel diodes, like used in amplifiers, this was in the tens of GHz.  Making a circuit that has this characteristic over that extreamly wide frequency range requires a lossy coax or lossy radial line.  You can just put a resistor in parallel with the TD and this will stabilize a little of the IV curve just past the peak and valley points.  As the value of the resistor gets lower and lower the area stabilized increases.  A short will completely stabilize the diode, but then you can't see the I-V curve.

Early tunnel diodes were made by hand one at a time.  A small sphere of metal was alloyed to a chip of semiconductor material (Germanium, Gallium arsenide, or Gallium Antimonide) The chip was die attached into a metal ceramic package and then had a gold screen bonded from one side of the package flange to the metal sphere and to the other side of the package, like a bridge.  Then the diode was etched in a caustic solution forming a mushroom shaped cross section.  The stem of the mushroom was extremely narrow and the actual junction was in this stem.  An improvement to that process involved gluing a couple of very small diameter glass rods (made by pulling a hot glass rod) on either side of the metal ball after sintering and before attaching the mesh.  The rods tended to take some of the stress off of the tunnel junction and make the diode much more reliable. The coefficient of thermal expansion of the glass was chosen to be close to the material of the tunnel diode.

Later a "planar process" was developed that allowed back diodes to be made at the wafer level.  This is used at Metelics.

Note that when you let the smoke out of a semiconductor it fails to proceed.  During the manufacturing processing, where the smoke is put into the semiconductor it is very important to know the lowest temperature that has an effect on the semiconductor proprieties.  In the case of tunnel/back diodes that temperature is much lower than for other diodes like PINs.  This means that the maximum operational temperature for tunnel/back diodes is much lower than for all other types of diode.

TRW bought Aertech because we were making amplifiers for classified space programs and they wanted more control.  Note that a TD amplifier draws a few mA of current at just over a volt.  We once got a return after maybe 5 years and the problem was that the Mercury "D" battery had died.  Tunnel and back diodes are heavily doped semiconductors.  This means that they are not effected by radiation.  Diodes like PINs that have light doping are very sensitive to an extra carrier and are used as the sensor in some radiation detectors.

Back Diode detectors have wide video bandwidths because the source impedance is low (a good match to a 50 Ohm IF system and well as to a 50 Ohm RF input).    They also have low flicker and 1/f noise.  We built a lot of detectors for military crystal video ECM receivers (Radar Warning Receivers).  They also make good mixers, just use a coupler to combine the RF and LO and feed it into a BD detector then into an IF strip for a quick noise figure measurement.

3626257 (Google) Semiconductor Device with Superlattice Region Dec. 7 1971 by Leo Esaki,
classes: 257/15; 148/DIG65; 148/DIG67; 148/DIG72; 148/DIG97; 148/DIG169; 257/1; 257/28; 257/E29.073; 257/E29.078; 257/E47.004  is probably the first Tunnel Diode patent and is the oldest patent in the 257/15 class.

4198644 (Google) Tunnel diode by Esaki; Leo 1980
TRW had a trade secret method of making the planar back diode and Metelics developed their own process.
The GE Transistor manual had a good section on their through hole Tunnel Diodes with circuit examples.
Note the GE tunnels are in a top hat type package that has leads, but the flange is needed to bond prior to etching.

Microwave Mixers

Bob Mouw was the first person to make a double balanced mixer that worked at microwave frequencies.  Prior to his invention double balanced mixers were made using a couple of ferrite balun transformers and a "ring" connected diode quad.  Their upper frequency limit was around 2 GHz.  Bob took the classical circuit and made a "dual circuit" that used a "star" diode quad and hybrid coaxial transmission line transformers.  I helped do the mathematical analysis for this mixer.  You can consider the diodes as switches that are turned on and off by the Local Oscillator.  The two states either pass the input signal or invert the input signal.  Doing an FFT (Wiki) on the waveform gives all the frequency domain outputs of a mixer, less those that are cancelled due to the mixer balance.

3512090 (Google) 5 /1970 HYBRID JUNCTIONS AND MIXERS AND MODULATORS CONSTRUCTED THEREFROM
Issued/ Filed: May 12, 1970 / Oct. 18, 1967 455/326; 333/24R; 455/331 - Is the first version of the Mouw mixer pattent
 
3818385 (Google) 06/18/1974 HYBRID JUNCTION AND MIXER OR MODULATOR 333/26; 333/35; 333/238; 333/243; 455/326
The diodes shown on the first page are in ceramic packages.  Later there were much more advanced versions built.  These typically were made in octave bandwidths. 3638126 (Google) 01/25/1972 High-frequency converter - Bob later worked for Spacek

Links

Unusual Diode FAQ -

Back to Brooke's Rack and Stack Web page created 2000