July 2012 Archives

Multi-level Doherty Amplifier

US Patent 6,700,444. N-way Doherty amplifier.

This is some neat stuff. While it can introduce some non-linearities to the resulting signal, this might be the secret to doing some neat stuff using multiple Class-C amplifiers. Specifically, I'm thinking about some of the older Motorola and GE amplifiers where they often used three or four amplifier bricks in parallel to achieve a given wattage out. To modify one of those, one would change the bias for each individual brick.


Now I wonder how a Class-C amplifier works in "overload". Obviously, if you've got one operating from say 5-55W and one from 55W-110W, you've got two ranges covered, so you should be able to produce power from 5-110W, continuously variable. Adding 0-5W would be trivial, but the input signal would have to be reduced to prevent overloads -- say by using a -10dB hybrid. The input signal would routed so that 90% of the power goes to the high-power amplifier input, and 10% of the power goes to the low power amplifier input. Well, your lower stages will go into clipping, but that isn't exactly the problem since the amp next power level up can "peak" when the lower one "clips".

NASA has done some work on unequal combining: Ka-Band Waveguide Hybrid Combiner for MMIC Amplifiers With. Unequal and Arbitrary Power Output Ratio. Simons, et. al.

Three Components of Amplification: Phase, Gain, and Distortion.

Phase adjustments are commonly made in the process of unequal combining, equal combining, and equalization to cancel out non-linearity. All corrections may be thought of as phase corrections, though implementation varies from one technique to another.

Transmitter linearity depends on the crossover point being controlled minutely and closely, either in the amplifier assembly itself or when combining amplifiers together.  The less an amplifier conducts, the more efficient it is. The more a transistor (or tube) is used in a switch-like manner -- either on or off -- the more efficient it is. In Class-A, Class-AB, and Class B amplifiers, it may be useful to think of the transistor as a variable resistor. A Class-C amplifier may be thought of as one half of a push-pull Class-B amplifier. A Class-D amplifier is a digital amplifier where the duty cycle of a switching transistor is varied according to the magnitude of output power required. This is commonly called pulse-width modulation (PWM). An Inductive Output Tube (IOT) may be thought of as elements of several amplifiers working together; indeed, it is. At various points of a NTSC television signal, the IOT acts as either a linear amplifier or a Class-C amplifier / RADAR magnetron. (With "digital television" becoming the dominant mode of broadcasting, the IOT is relegated to linear amplifier duty, at a derated power output rating from the rating used in NTSC service).

Too little gain results in poor efficiency; too much gain results in self-oscillation and spurious emissions.  Self-oscillation may be destructive in certain amplifiers. In RF amplifiers, it results in spurious emissions, mixing products, and/or excessive harmonics.

Amplifier Conduction Angles: Determining Amplifier Class Between Zero and 360 Degrees of a Sine Wave Carrier

In Class-A amplifiers, the amplifier conducts 360-degrees of the cycle of a sine wave that it is designed to amplify. Bipolar transistors require about 0.7VDC to "turn on"; this is related to the turn-on voltage of diodes, and varies according to application, design and manufacture. The Class-A amplifier is biased in such a way that the transistor is always conducting some small amount of power. This current is reduced to the lowest practical level by the circuit designer and is called quiescent current. This constantly flowing current results in heating when no signal is being amplified, and contributes negatively to efficiency. The transistor, constantly conducting even when quiet, is used as a variable resistor and accordingly heats up as output circuitry handles power.

In Class-B amplifiers, crossover distortion results when the signal switches between the turn-on point of two amplifiers which are configured to work on the opposite phase of each other. To put it another way, one half of the Class-B amplifier works on the negative part of the signal, the other on the positive. Bipolar transistors require about 0.7VDC to "turn on"; this is related to the turn-on voltage of diodes. Since diode (and transistor construction) vary by material and manufacture, this voltage may be lower or higher. Typically, Class-B amplifiers are built using matching or complimentary parts. When amplifying a sine wave, each half of the amplifier conducts about 180 degrees -- about, but not quite 180-degrees because of the turn-on voltage issue -- of the input signal, turning off for the other half of the cycle.

A Class-C amplifier is but one half of a Class B or push-pull amplifier, conducting less than 180 degrees for the entire cycle. Typically, the rest of the cycle is made up from a resonant "tank" circuit "ringing" or resonating slightly -- as if a glass might resonant after struck by a finger.

A Class-D amplifier operates like a

A Class-D amplifier is a digital amplifier where the duty cycle of a switching transistor is varied according to the magnitude of output power required. This is commonly called pulse-width modulation (PWM).

Modern transmitters use pre-distortion to manage crossover (voltage) distortion; they take advantage of the non-linearities of the amplifier, sample the result and calculate and generate an error signal. The error signal is then added to the input signal resulting in cleaner output after amplification.

The way this actually works in the analog and pre-digital or pre-DSP world is that the signal is distorted (squared, cubed, or raised-cosine) and the resulting harmonics are selectively applied to mostly correct the non-linearity without introducing more artifacts.  In the digital world, a complex mathematic number (a polynomial) is calculated from the sampled signal by a DSP or FPGA and sent to a DAC where it is turned into RF and combined with the input signal. So, in a way, two wrongs (the original non-linearity and the distorted product apply in an inverse fashion) make a right. Bell Labs experimented with this in various methods, Doherty, and so on. It is the basis of an equalizer, or an adaptive equalizer.

Bell had this figured out by 1940 by Bennett and used manually in L3, L4, L5 coaxial systems:

Bell System Technical Journal, v19: October 1940: Cross-Modulation Requirements on Multichannel Amplifiers Below Overload (PDF).

By the time microwave radio came into use, with various propagation phenomenon such as selective fading, etc. gumming up the works, it became necessary to make changes automatically. Bell people figured this out and a patent filed for June 8, 1973 and subsequently issued July 23, 1974: 3,825,843 (PDF). Other patents cite that patent that went on to evolve the mechanism until we get to today and DSP.

Interestingly, the patents don't seem to proliferate the technology as well the first decade or so. After the 23+ year period, people generally take the idea and apply it to everything. But I digress.

The caveat to multi-level amplification systems is that the resulting signals must be combined. Using a -3dB combiner, the crossover point has to be set so that the Class-C amp is used to maximum efficiency and the Class A or AB amp used at minimum.This gets complicated when a third amplifier is introduced, as the additional losses associated with that amplifier must be considered. Using our hypothetical example of a 55W-110W amplifier and it's close second 5.5W-11W amplifier, we discover that adding an additional branch may cause more trouble than it is worth. Crossover distortion causes trouble when combining dissimilar power amplifiers as well.

Consider the following:

The input signal is broken into three equal branches. Each of three amplifiers are biased at different levels, resulting in various conduction angles between the three of them.

Amplifier A is set at the lowest threshold, so this amplifier conducts almost 100% of the time (almost 360 degrees)
Amplifier B is set to conduct about 50% of the time (180 degrees).
Amplifier C is set to conduct 5% of the time (18 degrees).

Amplifier A will be continously in operation. Obviously, it is Class A, Class AB, or two halves of a Class B amplifier -- linear. Forward loss for it will always appear to be 0dB if the other two are operating at the same levels. If the other two are NOT operating, the apparent losses will be on the order of -6dB or so. This is because the hybrid combiner connecting all three amplifiers together imposes a loss in the forward signal if it is not in phase with the other signal. Additionally, if the second amplifier is not running, then the resulting power will be -3dB -- because one of the two amplifiers are running .

add section on breaking high power into two amplifiers, note that group delay increases.

Another approach to SDR

Earlier I wrote an article vilifying the authors of an article on Twist-Shift Keying (TSK), which is an attempt to modulate a signal in the polarization plane in a binary form. While there is no doubt that one could certainly do so, the method has not been used because the antenna matching is difficult. Cross-polarization loss from linear to circular polarizations is only -3dB (circular to linear or vice-versa). However, isolation (cross-polarization loss) between orthogonal modes of the same polarization mode (opposites of either linear/linear or circular/circular) is theoretically infinite, however practically reaches about -30dB.

However, thinking about SDR and methods to achieve power, I started thinking back to the Twist-Shift Keying (TSK) implementation method and realized that the parts themselves could be used to generate signals at power.

Traditionally, a SDR requires a linear amplification chain. Two-way radios do not typically have linear amplifier chains, as they are mostly FM and/or use FM-derived modulation methods. Accordingly, few radios have any linear stages in the transmitters, as Class C amplifiers provide about 80% efficiencies and linear amplifiers top out at around 50% unless modifications are made. (See Doherty.)

One of the basic problems with SDR is that QAM and QPSK modes cause modulation of the output power. This is not easy to achieve using a Class-C amplifier, as a Class-C amp can practically only handle so much modulation of the input power; speaking from a practical perspective, a Motorola Syntor X transmitter is only spec'd to run from 55W - 100/110W for the 100/110W radio and even then only certain models. However if one drives another radio at 180-degrees out from the first, the result is a null in power at the meeting-point, albeit at 100% power wasted. In other words, it's possible to use the amplifiers to create a linear amplifier, with all of the disadvantages of a linear amplifier.

However, working this backwards mentally, you arrive at the basic definition of RF as seen by an SDR: I & Q, or "In-Phase" and "Out of Phase". The further problem with this however is that when the I&Q are 180-degrees out and join, any slight offset (group delay) or (transmitter phase) noise will result in a generated sideband, spur or mixing product. Filtration requirements go up, waaay up. Think "those annoying high frequencies you might hear in your noise-cancelling headphones". Except all over the bands, and worse, at power levels large enough to concern the FCC. Think "transmitter mask filter" or "interdigital filter" and you're approaching the cost point -- which is why we don't see this approach used at this time.

So the basic approach is to use two DACs to generate a signal and combine them, then amplify. It may be possible to use two VCOs to accomplish the same goal, provided they can be phase-modulated in equal amounts from separate DACs at baseband. The Syntor X -- like the Maxtrac and many other radios -- generates a signal from the VCO on frequency, not offset.

However, if one wants to generate a signal at a lower frequency, like a reverse Tayloe-Detector (commutating modulator), some of those radios with doublers and tripplers make a bit more sense -- Like the GE Mastr II and the Motorola Mitrek.

SDR Resources


Finding HBA WWNs under Solaris 8, 9, 10

About this Archive

This page is an archive of entries from July 2012 listed from newest to oldest.

June 2012 is the previous archive.

August 2012 is the next archive.

Find recent content on the main index or look in the archives to find all content.